Caches for continuous benchmarking should be branch specific

[lorenzwalthert]

Otherwise, cache_applying takes long the first time to run (as the cache has to be built), and then is instant, but when you compute the average per branch, one will have the first run which will drastically increase the mean.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.46%. Comparing base (014d1c8) to head (04a4370).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #805   +/-   ##
=======================================
  Coverage   90.46%   90.46%           
=======================================
  Files          47       47           
  Lines        2433     2433           
=======================================
  Hits         2201     2201           
  Misses        232      232           

☔ 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.

[github-actions]

This is how benchmark results would change (along with a 95% confience interval in relative change) if 4e2d570 is merged into master:

• cache_applying: 0.05s -> 0.05s [-0.85%, +4.62%]
• cache_recording: 1.07s -> 1.07s [-1.33%, +1.73%]
• without_cache: 2.85s -> 2.86s [-0.94%, +1.39%]

Further explanation regarding interpretation and methodology can be found in the documentation.