feat(ci): use pytest-split for optimal test distribution

[vringar]

Summary

Replaces the alphabetical glob-based test splitting (test/test_[a-e]*, etc.) with pytest-split for duration-based optimal distribution across CI runners.

• Adds pytest-split dependency and a .test_durations file with real timing data from CI
• Updates workflow matrix from glob patterns to numbered groups (1-7)
• scripts/ci.sh conditionally uses --splits/--group in CI, falls back to $TESTS for local use

Current bottleneck: test/test_[p]* (mostly test_profile.py) takes ~31.9 min while other groups finish in 3-12 min.

Expected improvement: With optimal splitting, max runner time drops from ~31.9 min to ~9.5 min (3.4x speedup).

Supersedes #1103 which attempted the same goal with a custom bin-packing script and static distribution file. pytest-split is a well-maintained plugin that handles new/renamed tests gracefully (assigns default durations) and uses a proven least_duration splitting algorithm.

Test plan

• CI passes with all 7 test groups running via pytest-split
• Verify test groups are roughly balanced in duration (check CI job times)
• Confirm coverage upload still works (artifact names changed from strategy.job-index to matrix.group)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 56.64%. Comparing base (d327d69) to head (2a368f5).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1141   +/-   ##
=======================================
  Coverage   56.64%   56.64%           
=======================================
  Files          40       40           
  Lines        3910     3910           
=======================================
  Hits         2215     2215           
  Misses       1695     1695           

☔ 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.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[Copilot]

Pull request overview

This PR improves CI test runtime by replacing static glob-based splitting with pytest-split, using historical per-test durations to balance 7 parallel test groups more evenly.

Changes:

• Add pytest-split to the project environment and introduce a .test_durations dataset for duration-based scheduling.
• Update CI to run tests as 7 numbered groups (1–7) rather than filename glob buckets.
• Adjust scripts/ci.sh to use pytest-split when GROUP/SPLITS are set, while keeping a $TESTS-based local fallback.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
scripts/ci.sh	Switches between CI duration-based splitting and local $TESTS execution.
environment.yaml	Adds pytest-split dependency needed by CI.
.test_durations	Provides real timing data for pytest-split’s least_duration algorithm.
.github/workflows/run-tests.yaml	Replaces glob matrix with numeric groups and updates coverage artifact naming + failure summary step.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

In this embedded Python script, the group label is written as ${{ matrix.group }} inside a Python f-string. GitHub expressions are not evaluated inside the runtime script (and {{/}} are treated as literal braces by Python), so the step summary will show a literal ${ matrix.group } instead of the actual group number. Pass the matrix value into the step via env: (or substitute it in bash) and use that value when writing the summary.