Debounce AI hooks

[paulpetit-gg-ext]

Context

Some agents can detect multiple hook configurations (for instance VSCode can load Claude's hooks in addition to its own).

In these cases, the same payload is usually called once for each hook, which create unnecessary load to GIM and latency to the user.

What has been done

• Remember the hash of the latest payload in a cache file and use it to debounce hook payloads.
• To avoid a race condition where the two hook processes are spawned quickly enough for the cache to be stale, file locking is done.
• New dependency to the filelock library that makes this locking cross-platform.

[codecov]

Codecov Report

❌ Patch coverage is 90.00000% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 93.63%. Comparing base (befdca6) to head (87e9b29).

Files with missing lines	Patch %	Lines
ggshield/verticals/ai/hooks.py	90.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1252      +/-   ##
==========================================
- Coverage   93.64%   93.63%   -0.01%     
==========================================
  Files         183      183              
  Lines        9960     9980      +20     
==========================================
+ Hits         9327     9345      +18     
- Misses        633      635       +2     

Flag	Coverage Δ
unittests	93.63% <90.00%> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[xblanchot-gg]

A few small comments, overall looks good !