Stabilize mature interfaces and remove retired config

[max-sixty]

Marks the mature wt step eval, wt step for-each, wt step prune, LLM summary, config-state vars, and commit-template append interfaces as stable.

Removes compatibility for the long-deprecated template-file and squash-template-file settings, and retires the special migration for the already-ignored switch.picker.timeout-ms. These keys now follow the ordinary unknown-field path; the troubleshooting guide explains how to move file contents into inline templates. The change also removes the redundant project CI-platform accessor.

Validation: cargo run -- hook pre-merge --yes (4,710 tests passed; one skipped), plus a production documentation build.

This was written by Codex on behalf of @max-sixty

[worktrunk-bot]

The stabilization sweep looks complete on the generated pages, but the hand-edited llm-commits.md was missed — it's the primary for two of the six interfaces this PR promotes, and nothing in the sync test catches a badge in prose.

docs/src/content/docs/llm-commits.md still marks both features experimental. Under ## Branch summaries and under ## Appending to the prompt there's a <span class="badge-experimental"></span> on the line after each heading. Per docs/CLAUDE.md's sync taxonomy that file is primary, with the [experimental] copies in skills/worktrunk/reference/llm-commits.md and plugins/worktrunk/skills/worktrunk/reference/llm-commits.md generated from it — so the fix is to drop both spans and regenerate. As it stands, list.md describes LLM summaries as stable while the dedicated LLM page still badges them, and the same split applies to template-append, which the PR de-badges in config.md, dev/config.example.toml, dev/wt.example.toml, and the CommitGenerationConfig doc comment. Happy to push that commit if you want it.

The retired template keys land on a message that doesn't say what happened. A user with commit.generation.template-file now sees User config has unknown field commit.generation.template-file (will be ignored) and their commit prompt silently falls back to DEFAULT_TEMPLATE — the warning reads as a typo report rather than "the key you configured is gone and here's where the contents go". ListTaskTimeout kept a DeprecationKind for exactly this shape (list.task-timeout-ms is no longer used — list.timeout-ms bounds the collect phase), and an UpdateOnly row would warn with that guidance while keeping the test_warning_fires_iff_update_changes invariant intact. I can see the argument for the current choice — leaving the key in place preserves the path so the user can still find their template file — so this is a judgment call rather than a defect; flagging it because the troubleshooting page is the only place the migration path is written down and nothing in the warning points there.

Not approving: the diff edits the wt step prune declaration in src/cli/step.rs, promoting a bulk worktree-and-branch removal command out of experimental. Per .claude/skills/running-tend/references/review-pr.md → Data-Loss Surface: Hold for Human Review, changes on that surface aren't mine to sign off, and a stability promise on a bulk-delete command is the kind of signal that changes how readily users reach for it. Everything else here reads correct to me: exclusive_sibling and the mutual-exclusivity validation go together cleanly, ProjectConfig::ci_platform had no callers left outside its own test, and the retired switch.picker.timeout-ms rule is consistently gone from the rules table, the warning formatter, the combined snapshot, and both deprecation test batteries.