Sort surface_distance.py imports with isort

[brendancol]

Fixes #3710.

Runs isort over xrspatial/surface_distance.py so it passes the check the project already configures in setup.cfg (line_length = 100).

Categories addressed

Cat 4 (isort) only. Three things change:

1. from xrspatial.cost_distance import _heap_push, _heap_pop is alphabetized to _heap_pop, _heap_push.
2. The two separate from xrspatial.utils import ... statements, split by an unrelated dataset_support import, become one.
3. The xrspatial.utils block moves from a vertical-hanging-indent wrap to the grid wrap that cost_distance.py, proximity.py, focal.py and zonal.py already use.

No behaviour change is intended and none is possible here: the same names are imported from the same modules, only the order and the wrapping differ. Nothing was added, removed, or re-exported.

flake8 was already clean on this file (0 violations) and stays clean. Cat 1, 2, 3 and 5 turned up nothing to fix.

Verification

$ python -m isort --check-only xrspatial/surface_distance.py   # exit 0
$ python -m flake8 xrspatial/surface_distance.py               # exit 0
$ python -m pytest xrspatial/tests/test_surface_distance.py -q
37 passed in 3.72s

Run on a host with CUDA available, so the cupy and dask+cupy tests in that file ran for real rather than skipping.

Also in this PR

The .claude/sweep-style-state.csv row for surface_distance, including the two LOW findings recorded but deliberately not fixed (the dead height, width unpack in _surface_distance_dask, and _precompute_dd_grid reaching into xrspatial.zonal for a helper this module already defines). Both are described in #3710.