handle monotonicity of cftime objects

[larsbuntemeyer]

closes #581

Improves the handling of datetime timedelta objects in core dimension ordering and bounds monotonicity checks within the cf_xarray/helpers.py module. The changes ensure that comparisons involving timedeltas are correctly performed using their total seconds.

Enhancements for datetime timedelta support:

• Updated _get_core_dim_orders to detect and process datetime.timedelta differences using the total_seconds method for safe numeric comparison.
• Refactored _is_bounds_monotonic to check for datetime.timedelta in bounds differences and convert them to seconds before monotonicity evaluation, instead of always casting to float64.
• Add tests.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.64%. Comparing base (a9cebee) to head (3171796).
⚠️ Report is 67 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #582      +/-   ##
==========================================
+ Coverage   85.78%   86.64%   +0.86%     
==========================================
  Files          13       15       +2     
  Lines        2364     3033     +669     
  Branches      183      296     +113     
==========================================
+ Hits         2028     2628     +600     
- Misses        303      353      +50     
- Partials       33       52      +19     

Flag	Coverage Δ
mypy	44.26% <11.11%> (+5.73%)	⬆️
unittests	93.52% <100.00%> (-0.47%)	⬇️

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.

[aulemahal]

Looks good! Thanks for the fix, and sorry for the breaking change. I didn't think of time dtypes when reviewing this.

Out of curiosity, what workflow are you doing that requires a bounds_to_vertices on the time bounds ?

I can approve and release the patch when a test is added.

[larsbuntemeyer]

Out of curiosity, what workflow are you doing that requires a bounds_to_vertices on the time bounds ?

I do lots of cmorization! Sometimes need it for the cmor.axis API, but now as you asked, i looked closer into the cell_bounds keyword and it seems it also accepts the 2d version with (2, n) dims.

Yes, i need it for cmorization

In the case of a 1-d array, the size is one more than the size of coord_vals and the cells must be contiguous.