Handle empty sequences in convert_ras_hdf_value

[arpitjain099]

convert_ras_hdf_value() in src/rashdf/utils.py raises IndexError when it is
handed an empty sequence. The list/tuple/ndarray branch only distinguishes
len(value) > 1 from everything else, so an empty value falls into the
single-element else branch and indexes value[0]:

elif isinstance(value, (list, tuple, np.ndarray)):
    if len(value) > 1:
        return [convert_ras_hdf_value(v) for v in value]
    else:
        return convert_ras_hdf_value(value[0])  # value[0] -> IndexError when empty

An empty array is easy to hit when an HDF attribute or dataset carries no
elements, and the resulting IndexError is harder to trace back than a clean
return. This adds an explicit len(value) == 0 case that returns None, which
matches how the function already maps absent values (NaN becomes None) and
stays inside the declared return type.

Verification with pytest on the affected test:

• Before the fix, tests/test_utils.py::test_convert_ras_hdf_value fails with
  IndexError: list index out of range at src/rashdf/utils.py:295.
• After the fix it passes, and the full tests/test_utils.py module is green
  (6 passed).

The test adds coverage for the empty list, tuple, and numpy array inputs.

I hit some unrelated failures in test_geom.py and test_plan.py in my local
environment, but they reproduce on a clean checkout of main as well (they look
like geopandas/zarr version differences), so they are not related to this
change.