fix: address prior PR feedback — CI mutation grep, docs, eval DAG

[haasonsaas]

Summary

• CI (PR40 feedback): Fix mutation job grep so it extracts the missed count (e.g. 2 from 14 mutants ... 2 missed, 9 caught), not the first number (14). The previous grep -oE '[0-9]+' | head -1 picked total mutants and caused spurious baseline failures.
• README: Add PR workflow bullet under Contributing (CI green, test plan, small PRs).
• docs/mutation-testing: State that *state* and *storage_pg* are not run in CI; add table from local run and re-run commands.
• eval DAG (PR39 feedback): Add comment that ExpectationMatching / CommentCountValidation / BenchmarkMetrics form a linear chain so parallelizable: true has no effect today; keep for consistency.

Test plan

• cargo test (no code behavior change beyond comment).
• CI mutation job should now compare correct missed count to baseline.

Made with Cursor

[cursor]

PR Summary

Low Risk
Low risk: changes are limited to CI log parsing plus documentation/comment updates, with no runtime behavior changes in the shipped code.

Overview
Fixes the CI mutation-testing gate to extract the actual "missed" mutant count from cargo mutants output (avoiding false failures caused by capturing the total mutant count).

Updates contributor docs to clarify the PR workflow and to document that mutation testing for *state*/*storage_pg* is local-only (with a brief local-run summary and rerun commands). Adds a clarifying comment in the eval fixture DAG noting that several stages currently form a linear chain so parallelizable: true has no practical effect today.

^{Written by Cursor Bugbot for commit 39e5f9b. This will update automatically on new commits. Configure here.}