docs(classify): fix stale example outputs and add missing examples

[brendancol]

Fixes the documentation issues found by the documentation sweep on the classify module (#3506).

Stale / wrong example output

• reclassify: the numpy, dask, and cupy example blocks all printed an array that does not match the function's result for the example input. Corrected all three to the actual output.
• equal_interval: the example printed a stale array. Corrected to the actual output.
• binary: the example used np.nan in the array repr where numpy prints nan. Fixed.

test_classify.py fixtures confirm the functions return the corrected values, so the code was right and only the docstrings were stale.

Missing Examples

Added an Examples section to std_mean, head_tail_breaks, percentiles, maximum_breaks, and box_plot. Each output was generated by running the function and pasted verbatim.

Scope

Docs only. No behavior change. test_classify.py passes (102) and flake8 is clean.

The natural_breaks num_sample/None type omission (already tracked in #3501) is intentionally left out of this PR.

[brendancol]

Review: classify docstring examples

Re-ran every docstring example on a CUDA host and diffed the printed output against actual output.

Blockers: none.

Suggestion (fixed in ef98faa): the reclassify docstring had its three result arrays corrected by the PR, but the surrounding lazy reprs in the same docstring were still stale:

• the agg_da repr showed dtype=float32, but the float64 input gives dtype=float64
• the lazy result repr showed dask.array<_run_numpy_bin, ...>; the current task name is dask.array<xrspatial.reclassify, ...> after the _dask_task_name_kwargs rename
• the cupy block printed <class 'cupy.core.core.ndarray'>; current cupy returns <class 'cupy.ndarray'>

All three corrected against live output.

Nit (left as-is): the example headers omit the Size: NNB suffix modern xarray prints. Kept consistent with the other examples in this module (quantile, natural_breaks); these are .. sourcecode:: blocks, not doctests.

What looks good: the two genuinely wrong outputs (reclassify, equal_interval) now match actual output; the five new Examples print exactly what the functions produce; docs-only with no behavior change; 102 tests pass; flake8 clean.