Fix polyval returning a sentinel value for NaT in timedelta coordinates

[dchaudhari7177]

Fixes#11462.

_ensure_numeric (used by polyval and polyfit) converts a timedelta64 coordinate with a plain astype(float). NaT is stored as the int64 minimum, so that cast turned it into a large sentinel (-1.84e19) rather than propagating a missing value.

Fix: mask the NaT positions back to NaN after the cast (.where(x.notnull())), mirroring the datetime_to_numeric path used for the datetime branch. Valid values keep their existing numeric scale, so only the previously-bogus NaT output changes.

MCVE from the issue now returns [1.0, nan, 345601.0] as expected. Added test_polyval_timedelta_nat (fails on main, passes with the fix); the existing polyval tests stay green. Added a whats-new.rst entry.

• Closesxr.polyval returns bogus values for NaT in timedelta arrays #11462
• Tests added
• User visible changes (including notable bug fixes) are documented in whats-new.rst

[welcome]

Thank you for opening this pull request! It may take us a few days to respond here, so thank you for being patient.
If you have questions, some answers may be found in our contributing guidelines.

[welcome]

Congratulations on completing your first pull request! Welcome to Xarray! We are proud of you, and hope to see you again!