fix(compression): avoid regex for repeated literal blocks

[Jehu]

Summary

The session-dedup compressor now counts repeated text blocks with literal string scanning instead of compiling each block as a regular expression. Large memory or JSON-like blocks can no longer trip V8's Regular expression too large limit while the existing dedup behavior stays intact.

Test plan

• DISABLE_SQLITE_AUTO_BACKUP=true node --max-old-space-size=8192 --import tsx/esm --import ./open-sse/utils/setupPolyfill.ts --import ./tests/_setup/isolateDataDir.ts --test --test-force-exit tests/unit/compression/session-dedup.test.ts tests/unit/compression/session-dedup-memory-7849.test.ts tests/unit/compression/session-dedup-intra-message-6467.test.ts

[diegosouzapw]

Nice fix — this lines up with the project's existing regex-size guardrails, and the guarded-RegExp test (asserting no RegExp is ever compiled from the block) is a clean way to prove it structurally. Before merging, could you add a changelog fragment under changelog.d/fixes/ referencing #6467? Everything else looks solid — the literal-scan replacement functions correctly preserve the "keep the first occurrence" behavior of the original regex-based code.

[Jehu]

could you add a changelog fragment under changelog.d/fixes/ referencing #6467?

Added in changelog.d/fixes/14633-session-dedup-literal-scan.md and pushed in b9e5dda.