test(integration): add pipeline stage benchmarks and non-mutating lint format check

[HarshK97]

Problem

1. We had no benchmarks for individual engine stages like parse, match, edit script, postprocess, and serialize, so tracking performance or memory regressions per stage was impossible.
2. CI had no automated benchmark regression checks or A/B diff comparisons against main to catch performance slowdowns on PRs.
3. Running make lint formatted code directly on disk, which mutated files in place instead of catching unformatted code and failing.

What Changed

• Pipeline Stage Benchmarks: Added benchmarks for parse, match, edit script generation, postprocess, serialize, and full pipeline runs across all test fixtures.
• CI Benchmarking & A/B Reports: Added a benchmark smoke test and automated benchstat comparison against the base branch on ubuntu-latest, posting the comparison table directly to the GitHub Actions job summary.
• Makefile Commands: Added make bench and make bench-short. Fixed make lint to run a non-mutating format check that prints a diff and fails if formatting errors exist.
• Benchmark Test Helpers: Swapped *testing.T to testing.TB on test helpers so fixture loading and parsing are shared between standard integration tests and benchmark loops.