Using np.nextafter instead of eps

[yfukai]

This pull request updates the tile slicing logic in the _yield_apply_tiled function within src/tracksdata/functional/_apply.py to improve numerical precision and ensure correct handling of inclusive and exclusive ranges. The most important changes are:

Tile slicing precision improvements:

• Replaced the use of a fixed epsilon (eps) for adjusting slice end points with np.nextafter, which more accurately converts inclusive to exclusive ranges based on floating-point precision. This change affects both the calculation of slicing_without_overlap and the main slicing logic.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 88.22%. Comparing base (cb4770a) to head (40da2ae).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #189      +/-   ##
==========================================
- Coverage   88.22%   88.22%   -0.01%     
==========================================
  Files          51       51              
  Lines        3593     3592       -1     
  Branches      617      617              
==========================================
- Hits         3170     3169       -1     
  Misses        254      254              
  Partials      169      169              

☔ 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.

[yfukai]

Now it's ready to review @JoOkuma! The edge case fails with the previous impl but succeeds with np.nextafter.

[JoOkuma]

LGTM, @yfukai

I didn't think of using -np.inf. Thank you for the fix and for implementing the edge case.