Style cleanup in viewshed.py: rename shadowed id, fix E127 + isort

[brendancol]

Closes #2690.

Style-only cleanup of xrspatial/viewshed.py. Limited to what flake8 (max-line-length=100) and isort (line_length=100) flag, plus one manual rename. No formatter was run and no config was widened.

What changed:

• Renamed the local id to node_id at the two _insert_into_tree call sites (lines 1409 and 1474). It was shadowing the id builtin, and the value already maps to the callee's node_id parameter, so the name now matches. Pure local rename.
• Fixed E127 over-indented continuation lines in the _viewshed_distance_sweep signature.
• Reflowed the .utils import block for isort.

This covers the style sweep's Cat 1 (E127), Cat 4 (isort), and Cat 5 (shadowed id builtin). None of it changes runtime behavior; the id rename is just a local variable inside numba-jitted code, and the rest is formatting. Style fixes apply uniformly across the numpy/cupy/dask backends, so there's no per-backend verification to do.

Test plan:

• flake8 xrspatial/viewshed.py reports nothing
• isort --check-only --diff xrspatial/viewshed.py is clean
• the 25 tests in xrspatial/tests/test_viewshed.py pass
• a numpy viewshed smoke test exercises both renamed _insert_into_tree call sites

[brendancol]

Review: style cleanup in viewshed.py

I read the full diff plus the surrounding code in _viewshed_cpu_sweep and _viewshed_distance_sweep, and re-ran the tooling against the head branch.

Blockers: none. Suggestions: none. Nits: none.

What looks good:

• The id to node_id rename is complete. No bare id tokens are left in the file, and node_id is already the parameter name in _insert_into_tree (line 321) and _find_value_min_value (line 75), so the rename follows existing module naming rather than inventing something new.
• No collision: node_id wasn't a local in _viewshed_cpu_sweep before, so nothing gets clobbered.
• The E127 fix lines the continuation up with the opening paren in the _viewshed_distance_sweep signature. flake8 reports zero now.
• The isort change just rewraps the .utils block; the imported names are identical.
• Checked locally: flake8 exits 0, isort --check-only exits 0.
• Behavior is unchanged. The renamed locals sit inside numba-jitted code and a local rename doesn't affect codegen. The 25 tests in test_viewshed.py pass and a numpy smoke test runs through both rename sites.

Checklist (most items not applicable for a static style change):

• Algorithm vs reference: n/a, no algorithm change
• Backend consistency: n/a, applies uniformly
• NaN handling: n/a, unchanged
• Edge-case tests: existing coverage unchanged and passing
• Dask chunk boundaries: n/a, unchanged
• Premature materialization: n/a, unchanged
• Benchmark: not needed
• README feature matrix: n/a, no API change
• Docstrings: unchanged