aspect: fix flake8 E402/E305/E501 in aspect.py

[brendancol]

Closes #2683

What

• Moved the from xrspatial.geodesic import (...) block from below _geodesic_cuda_dims up to the other top-of-file imports. Fixes E402 (import not at top) and E305 (blank lines after function def).
• Wrapped two over-long _run_gpu_geodesic_aspect[griddim, blockdim](...) kernel-launch calls. Fixes both E501 (101 and 109 chars).

flake8 xrspatial/aspect.py now reports zero violations.

Categories addressed

Cat 1 (flake8 E-codes: E402, E305, E501).

Cat 4 (isort) was reviewed and intentionally not applied: slope.py and curvature.py use one-import-per-line for xrspatial.utils, and isort --check-only flags them the same way. Applying the raw isort diff would collapse aspect's imports and make it inconsistent with its sibling terrain modules. No config widening, no # noqa.

Cat 2, Cat 3, and Cat 5 (bare except / mutable defaults / == None / shadowed builtins) were grepped and are clean.

No behavioural change

Static reformatting only. All four backends (numpy / cupy / dask+numpy / dask+cupy) share the same source, so no backend-specific verification is needed.

Test plan

• flake8 xrspatial/aspect.py clean
• pytest xrspatial/tests/test_aspect.py (64 passed)
• pytest xrspatial/tests/test_geodesic_aspect.py (18 passed)
• Import + smoke test of aspect / northness / eastness

[brendancol]

PR Review: aspect: fix flake8 E402/E305/E501 in aspect.py

Blockers

None.

Suggestions

None.

Nits

None.

What looks good

• The diff is exactly the three intended changes and nothing else. No adjacent code touched.
• Relocating the xrspatial.geodesic import to the top is semantically identical. The module imports cleanly with no dependency on _geodesic_cuda_dims, confirmed by the passing test run. This also resolves the real readability problem of an import hidden below a function def.
• Both line-wraps are whitespace-only inside the argument lists of the _run_gpu_geodesic_aspect[...] kernel launches. No behavioural change.
• The decision to skip the isort reordering is correct: slope.py and curvature.py use one-import-per-line for xrspatial.utils, so applying the raw isort diff would make aspect the odd one out. Documented in the PR body rather than silently suppressed.

Checklist

• Algorithm matches reference/paper (no algorithm change)
• All implemented backends produce consistent results (static reformat; shared source)
• NaN handling is correct (untouched)
• Edge cases are covered by tests (existing suite exercises every changed line)
• Dask chunk boundaries handled correctly (untouched)
• No premature materialization or unnecessary copies (no compute change)
• Benchmark exists or is not needed (not needed for a style fix)
• README feature matrix updated (not applicable)
• Docstrings present and accurate (untouched)

Verified locally: flake8 xrspatial/aspect.py clean; pytest test_aspect.py 64 passed; pytest test_geodesic_aspect.py 18 passed.