proximity: style cleanup (flake8 F841/E128, isort, mutable default)

[brendancol]

Closes #2725

Style sweep cleanup for xrspatial/proximity.py. Four findings from flake8 (max-line-length=100) and isort (line_length=100), bundled into one pass.

What changed

• Cat 3 (F841): removed the dead local original_chunks in the unbounded dask+cupy branch. It was assigned and never read, a leftover from when chunk geometry used to be restored after the cupy-to-numpy conversion.
• Cat 5 (mutable default): proximity, allocation, and direction declared target_values: list = []. Switched to a None sentinel and normalize to [] at the top of each function body. The old default was never mutated, so this preserves behaviour while dropping the shared-mutable-default anti-pattern.
• Cat 1 (E128): reflowed the under-indented np.where continuation in _vectorized_calc_direction.
• Cat 4 (isort): re-sorted the xrspatial import block and added the blank line isort wants after the inline import cupy as cp.

After the change, flake8 xrspatial/proximity.py and isort --check-only xrspatial/proximity.py both pass clean.

Behaviour

No behavioural change intended. The E128 reflow and isort reordering are formatting only, the F841 removal deletes dead code, and the mutable-default fix is behaviour-preserving (None normalizes to []).

Backends

Static source-level changes, uniform across numpy / cupy / dask+numpy / dask+cupy. No backend-specific logic touched.

Test plan

• flake8 xrspatial/proximity.py clean
• isort --check-only xrspatial/proximity.py clean
• Existing suite: pytest xrspatial/tests/test_proximity.py (69 passed)
• Added regression test: None default matches explicit [] for all three public functions

Note: test_proximity.py carries two pre-existing E127 warnings (lines 726, 752) and isort import drift that predate this PR. Left untouched to keep the diff scoped to the audited module.

[brendancol]

PR Review: proximity style cleanup (flake8 F841/E128, isort, mutable default)

Blockers (must fix before merge)

None.

Suggestions (should fix, not blocking)

None.

Nits (optional improvements)

• The annotation is now target_values: list = None, which really describes an Optional[list]. The strictly correct form would be Optional[list], but list = None is common shorthand in this codebase and the body normalizes immediately, so it is fine as-is. Not worth changing unless the project decides to standardize on Optional.

What looks good

• The F841 removal is safe: original_chunks had a single occurrence and was never read.
• The mutable-default fix preserves behaviour. The old [] default was never mutated, and None normalizes to [] at the top of each public function before any use.
• The E128 reflow and isort reordering are formatting only, confirmed clean by both linters.
• The regression test covers all three public functions via parametrize and asserts the None default matches an explicit empty list.
• The diff stays scoped to the audited module. Pre-existing lint in the test file was correctly left alone.

Checklist

• Algorithm unchanged (no logic touched)
• All backends: static source change, uniform across numpy/cupy/dask
• NaN handling unchanged
• Edge cases: existing 69 tests pass, new regression test added
• Dask chunk boundaries not touched
• No premature materialization introduced
• Benchmark not needed (no perf change)
• README feature matrix not applicable (no new function)
• Docstrings present and unchanged