Fixes #790: add .xrs xarray accessors

[brendancol]

Summary

• Adds .xrs DataArray and Dataset accessors so spatial operations can be called directly via tab-completion, aligning with the xarray ecosystem pattern (rioxarray .rio, etc.)
• Registers accessors automatically on import xrspatial — no extra setup needed
• Updates all user guide and example notebooks to use the accessor pattern, replacing per-function imports with the more convenient raster.xrs.slope() style

New files

• xrspatial/accessor.py — XrsSpatialDataArrayAccessor and XrsSpatialDatasetAccessor classes with lazy imports in every method body to avoid loading numba/cupy/dask at registration time
• xrspatial/tests/test_accessor.py — 21 tests covering registration, tab-completion, DataArray single/multi-input, and Dataset single/multi-input equivalence

API examples

import xrspatial

# Surface
raster.xrs.slope()
raster.xrs.hillshade(azimuth=315)

# Classification
raster.xrs.equal_interval(k=5)

# Multispectral — self is the NIR band
nir.xrs.ndvi(red)
nir.xrs.evi(red, blue)

# Dataset — single-input applies to each variable
ds.xrs.slope()

# Dataset — multispectral with band mapping
ds.xrs.ndvi(nir='band_8', red='band_4')

Notebooks updated

All 11 user guide and example notebooks converted to accessor style:
1_Surface, 2_Proximity, 3_Zonal, 4_Focal, 5_Classification, 6_Remote_Sensing, 7_Pathfinding, 9_Cost_Distance, zonal_crosstab, pharmacy-deserts, classification-methods

Test plan

• pytest xrspatial/tests/test_accessor.py -v — 21/21 passed
• pytest xrspatial/tests/ -v — 635 passed, 12 skipped, 0 failures
• Smoke test covering accessor patterns used in notebooks
• Verify notebooks run end-to-end (requires example data)

Closes #790