perlin: document name param, float-dtype requirement, and ValueError

[brendancol]

Documentation-only fix for perlin() docstring drift found by the documentation sweep.

The docstring listed only agg, freq, and seed in Parameters, but the signature is perlin(agg, freq=(1, 1), seed=5, name='perlin'). It also said nothing about the float-dtype requirement the function enforces.

• Documents the name parameter (sets the output DataArray's name).
• Notes on agg that input must be a floating-point dtype.
• Adds a Raises section for the ValueError on non-float input. This guards the regression from perlin() silently corrupts int-dtyped input rasters (all pixels -> INT_MIN) #1232 (integer input corrupted every pixel to INT_MIN) and is covered by test_perlin_rejects_integer_dtype.

No behavior change. Only the docstring was edited.

Backends: not applicable (no code path touched; all four backends unchanged).

Test plan:

• pytest xrspatial/tests/test_perlin.py -> 12 passed
• docstring example runs and reproduces the documented output array
• every signature parameter except agg now appears in Parameters; Raises section present

Note: the upstream repo has GitHub issues disabled, so there is no linked issue number.

[brendancol]

PR Review: perlin docstring fix

Documentation-only change. The diff edits the perlin() docstring and adds a sweep-state CSV row. No source behavior changes.

Blockers

None.

Suggestions

None.

Nits

None.

What looks good

• The new Parameters entry for name matches the signature default ('perlin') and the actual use at the return site.
• The float-dtype note and Raises section match the real guard in perlin() and the contract codified by test_perlin_rejects_integer_dtype (regression from #1232).
• The docstring example still reproduces its documented output array; the perlin test suite passes (12/12).
• Every signature parameter except agg now appears in Parameters.

Checklist

• Algorithm matches reference/paper: n/a (no code change)
• All implemented backends produce consistent results: n/a (no code change)
• NaN handling correct: n/a
• Edge cases covered by tests: existing coverage unchanged and passing
• Dask chunk boundaries handled correctly: n/a
• No premature materialization: n/a
• Benchmark exists or not needed: not needed
• README feature matrix updated: not applicable (no new function, no backend change)
• Docstrings present and accurate: yes, this PR fixes the drift