fix(ci): gate PR changes and add PR smoke-test for mutation-testing

[kaihowl]

Summary

• Removed hermetic git environment variables from mutation-testing workflow job
• Added pull_request path filter to run the workflow only when this workflow file changes
• Added PR-based quick smoke test for PR validation: on PRs, run a targeted mutation test; on schedule, run full mutation testing
• Test infrastructure already sets hermetic git settings automatically
• No other workflow changes besides env-var removal and PR-path gating

Problem

The weekly mutation testing workflow checkout could fail at the checkout step due to hermetic git environment variables being set at the job level, which prevented actions/checkout from configuring git's safe.directory.

Root Cause

These environment variables were redundant—the test infrastructure already configures hermetic git settings in tests:

1. Rust unit tests: hermetic_git_env() in git_perf/src/test_helpers.rs:20-26
2. Bash integration tests: export when spawning bash processes in git_perf/tests/bash_tests.rs:26-37
3. Bash test scripts: exported in test/common.sh:41-47

Solution

• Remove the hermetic git environment variables from the mutation-testing workflow:
  • GIT_AUTHOR_NAME / GIT_AUTHOR_EMAIL
  • GIT_COMMITTER_NAME / GIT_COMMITTER_EMAIL
  • GIT_CONFIG_NOSYSTEM
  • GIT_CONFIG_GLOBAL
• This allows checkout and all workflow steps to use normal git configuration while tests continue to run in hermetic environment via test infrastructure.
• Added pull_request path filter so the workflow only runs when this workflow file changes
• Added a PR-triggered smoke-test: when the workflow is invoked by a PR, run a quick mutation test against a single file to validate workflow wiring; scheduled runs still perform full mutation testing

Test Plan

• Verify workflow syntax is valid
• Gate mutation-testing workflow to PR changes (triggered when this workflow file is modified)
• PR quick smoke-test runs for PR validation
• Full mutation testing runs on schedule
• Checkout step completes without errors
• Mutation testing runs with hermetic git config (from test infrastructure)

Fixes #611

🤖 Generated with Claude Code

📎 Task: https://www.terragonlabs.com/task/83b22e1e-62d7-46cd-aa3f-6e277c4bed37

[github-actions]

Performance Report

⏱ Performance Results

Audit Results

✅ 'add-benchmark'
z-score (stddev): ↓ 0.17
Head: μ: 21.5ms σ: N/A MAD: 0ns n: 1
Tail: μ: 21.7ms σ: 892.3µs MAD: 294.4µs n: 25
 [-11.56% – +7.17%] ▇▅▆▅▁▅▅▄█▅▇▇▅▆▅▅▅▅▅▅▅▆▅▅▁▅
✅ 'release-binary-size'
z-score (stddev): ↑ 0.70
Head: μ: 8.1MB σ: N/A MAD: 0B n: 1
Tail: μ: 8MB σ: 9.3kB MAD: 2.3kB n: 25
 [-0.52% – +0.03%] ██████████████▆▆▆▆▆▆▆▆▆▆▁█
✅ 'report'
z-score (stddev): ↑ 0.63
Head: μ: 28ms σ: N/A MAD: 0ns n: 1
Tail: μ: 26.3ms σ: 2.7ms MAD: 1.6ms n: 38
 [-34.42% – +10.87%] ▇▇▇▇▃▇▇▆█▇█▇▇▇▆▆▆▆▆▆▆▇▇▆▅▄▄▄▄▄▄▄▁▄▄▄▅▄▇
✅ 'report-benchmark'
z-score (stddev): ↓ 0.19
Head: μ: 28.2ms σ: N/A MAD: 0ns n: 1
Tail: μ: 28.4ms σ: 1.2ms MAD: 599.1µs n: 25
 [-9.44% – +10.98%] ▆▅▅▅▄▄▅▄█▄▆▆▅▆▃▃▄▄▃▄▃▄▄▃▁▄
✅ 'report-size'
z-score (stddev): ↑ 1.29
Head: μ: 11.6kB σ: N/A MAD: 0B n: 1
Tail: μ: 5.5kB σ: 4.7kB MAD: 62B n: 38
 [-3.18% – +496.20%] ██████████████▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁█
✅ 'report-size-benchmark'
z-score (stddev): →
Head: μ: 11.6kB σ: N/A MAD: 0B n: 1
Tail: μ: 11.6kB σ: 0B MAD: 0B n: 14
 [+0.00% – +0.00%] ▅▅▅▅▅▅▅▅▅▅▅▅▅▅▅

Measurement Storage Size

Live Measurement Size Report
============================

⚠️  Shallow clone detected - measurement counts may be incomplete (see FAQ)

Number of commits with measurements: 259
Total measurement data size (on-disk (compressed)): 298.4kB

Repository Statistics (for context):
-------------------------------------
  Loose objects: 0 (0B)
  Packed objects: 2115 (983kB)
  Total repository size: 983kB

Created by git-perf