chore: Upgrade Rust to latest stable, 1.96.1

[grahamking]

The usual container and doc updates, plus new lints:

• zip takes things that implement IntoIter so the .into_iter() calls were unnecessary.
• sort_by_key is typically more efficient than sort_by.
• Use checked_div removes the need to check divisor is not 0.
• Use while let .. instead of loop { let ..

---

Summary by CodeRabbit

• Chores

  • Updated the pinned Rust version across build environments and project tooling to a newer release.
  • Aligned container build settings and local development defaults with the same Rust version.

• Documentation

  • Refreshed setup and development docs to match the latest Rust toolchain version.
  • Updated test-running guidance to use the current pinned toolchain.

• Style

  • Made small code cleanup improvements in iteration and sorting patterns.
  • Simplified several calculations for clearer handling when counts are zero.

[github-actions]

🌿 Fern Docs Preview: https://nvidia-preview-019f38e4-3ea0-73eb-af00-d4a8d30c727f.docs.buildwithfern.com/dynamo/dev

[coderabbitai]

Walkthrough

This PR bumps the pinned Rust toolchain from 1.93.1 to 1.96.1 across container Dockerfiles, Makefiles, rust-toolchain.toml files, and documentation. It also applies a batch of clippy-suggested refactors: checked_div for safe division, sort_by_key replacing sort_by, zip without into_iter(), while-let loop rewrites, and a clippy allow attribute.

Changes

Rust Toolchain Bump

Layer / File(s)

Summary

Toolchain version references container/templates/dynamo_base.Dockerfile, deploy/inference-gateway/epp/Dockerfile, deploy/inference-gateway/epp/Makefile, deploy/inference-gateway/ext-proc/Dockerfile, deploy/inference-gateway/ext-proc/Makefile, rust-toolchain.toml, lib/runtime/examples/rust-toolchain.toml, docs/development/custom-containers.md, docs/development/unified-backends.md, lib/kvbm-engine/CLAUDE.md, tests/parity/README.md

Rust version references (RUST_VERSION, RUST_IMAGE, channel) updated from 1.93.1/1.93 to 1.96.1 consistently across build files, config, and docs.

Clippy-Driven Code Refactors

Layer / File(s)	Summary
Safe division via checked_div lib/kv-router/src/indexer/branch_sharded.rs, lib/kvbm-physical/src/layout/config.rs, lib/mocker/src/scheduler/sglang/core.rs, lib/runtime/src/pipeline/network/egress/tcp_client.rs	Manual zero-check division guards replaced with checked_div(...).unwrap_or(0) while preserving zero-fallback behavior.
Zip iteration simplified lib/bindings/kvbm/src/block_manager/vllm/connector/leader/slot.rs, lib/kvbm-engine/src/leader/session/initiator.rs, lib/kvbm-engine/src/worker/group/spmd.rs, lib/llm/src/block_manager/pool/managed/inactive.rs, lib/llm/src/preprocessor.rs, lib/mocker/src/scheduler/sglang/decode.rs	Explicit .into_iter() calls before zip removed, relying on implicit IntoIterator, preserving the same pairing/iteration semantics.
Sort refactor to sort_by_key lib/llm/src/protocols/openai/chat_completions/aggregator.rs, lib/llm/src/protocols/openai/completions/aggregator.rs	choices sorting changed from sort_by with an index comparator to sort_by_key(|a| a.index) in both aggregation logic and tests.
Loop and misc clippy fixes lib/mocker/src/replay/offline/single.rs, lib/kv-router/src/indexer/pruning.rs, lib/runtime/src/transports/nats.rs, lib/runtime/examples/system_metrics/tests/integration_test.rs	Loop/break patterns rewritten to while-let, next_expiries peek/pop logic separated in a locked scope, a clippy allow attribute added, and an unused reqwest import removed.

Estimated code review effort: 2 (Simple) | ~15 minutes

Suggested labels: dependencies, rust-toolchain, clippy

Suggested reviewers: Rust maintainers familiar with kv-router, kvbm-engine, and llm aggregator modules

Poem:
A rabbit hopped from '93 to '96,
No more if-checks, just checked_div tricks,
Sort by key, zip without the fuss,
While-let loops replace the old "just in case, break, and cuss".
Toolchains updated, clippy appeased—
This bunny's build is finally, cleanly, pleased. 🐇

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	The description covers overview/details, but it omits the required reviewer-start section and Related Issues/issue confirmation.	Add the missing "Where should reviewer start?" section and the required Related Issues section, including the linked issue or no-issue confirmation.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly summarizes the PR’s main change: upgrading Rust to 1.96.1.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

---

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (1)

lib/kv-router/src/indexer/pruning.rs (1)

579-616: 🩺 Stability & Availability | 🔴 Critical | 🏗️ Heavy lift

Self-deadlock in queue_due
expiries stays live for the whole loop iteration, so push_worker_expiry() can re-lock self.next_expiries while that mutex is still held and hang the pruning worker. Scope the heap lock to a smaller block before touching per-worker state or calling push_worker_expiry().

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@lib/kv-router/src/indexer/pruning.rs` around lines 579 - 616, `queue_due` can
self-deadlock because the `next_expiries` mutex stays held across the whole loop
iteration while `push_worker_expiry()` may try to lock it again. Narrow the
lifetime of the `expiries` guard in `queue_due` so the heap lock is released
before accessing `self.workers`, `state.pop_expired`, or calling
`push_worker_expiry()`, keeping only the minimal critical section around
peek/pop.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Outside diff comments:
In `@lib/kv-router/src/indexer/pruning.rs`:
- Around line 579-616: `queue_due` can self-deadlock because the `next_expiries`
mutex stays held across the whole loop iteration while `push_worker_expiry()`
may try to lock it again. Narrow the lifetime of the `expiries` guard in
`queue_due` so the heap lock is released before accessing `self.workers`,
`state.pop_expired`, or calling `push_worker_expiry()`, keeping only the minimal
critical section around peek/pop.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 4d6ded8f-bfe1-4a12-9deb-1e97bbb7e967

📥 Commits

Reviewing files that changed from the base of the PR and between b032d7a and 92805da.

📒 Files selected for processing (27)

• container/templates/dynamo_base.Dockerfile
• deploy/inference-gateway/epp/Dockerfile
• deploy/inference-gateway/epp/Makefile
• deploy/inference-gateway/ext-proc/Dockerfile
• deploy/inference-gateway/ext-proc/Makefile
• docs/development/custom-containers.md
• docs/development/unified-backends.md
• lib/bindings/kvbm/src/block_manager/vllm/connector/leader/slot.rs
• lib/kv-router/src/indexer/branch_sharded.rs
• lib/kv-router/src/indexer/pruning.rs
• lib/kvbm-engine/CLAUDE.md
• lib/kvbm-engine/src/leader/session/initiator.rs
• lib/kvbm-engine/src/worker/group/spmd.rs
• lib/kvbm-physical/src/layout/config.rs
• lib/llm/src/block_manager/pool/managed/inactive.rs
• lib/llm/src/preprocessor.rs
• lib/llm/src/protocols/openai/chat_completions/aggregator.rs
• lib/llm/src/protocols/openai/completions/aggregator.rs
• lib/mocker/src/replay/offline/single.rs
• lib/mocker/src/scheduler/sglang/core.rs
• lib/mocker/src/scheduler/sglang/decode.rs
• lib/runtime/examples/rust-toolchain.toml
• lib/runtime/examples/system_metrics/tests/integration_test.rs
• lib/runtime/src/pipeline/network/egress/tcp_client.rs
• lib/runtime/src/transports/nats.rs
• rust-toolchain.toml
• tests/parity/README.md

💤 Files with no reviewable changes (1)

• lib/runtime/examples/system_metrics/tests/integration_test.rs

[datadog-official]

🎯 Code Coverage (details)
• Patch Coverage: 100.00%
• Overall Coverage: 41.29% (-4.11%)

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: a137bda | Docs | Give us feedback!}