Fix failing tests

[kounelisagis]

We currently have two failing tests:

• test_take_current_domain_into_account_sparse_indexing_sc61914 - Changed to our own assertion function since using pd.DataFrame.equals was resulting in an error for windows with numpy 1 even though the Dataframes were equal.
  More: The nightly build with earliest supported numpy job failed on Friday (2025-02-14) #2161
• test_dask_multiattr_2d - used rerun_except="TileDBError" instead of needing to import error types from two different packages. This was leading to ModuleNotFoundError: No module named 'distributed' for Windows platforms. This was maybe caused by missing test dependencies, but we can avoid that.
  More: https://dev.azure.com/TileDB-Inc/CI/_build/results?buildId=41877&view=logs&j=ea6b747a-d678-5a49-8233-1092783ec463&t=1da0fa6e-ac3c-538d-9479-a05b27431549&l=1694

Daily tests run: https://github.com/TileDB-Inc/TileDB-Py/actions/runs/13333400928

[ihnorton]

Seems fine, but can you please add some information for future reference about why these fix the errors?

Changed to our own assertion function

Why? (Was this the source of the aborts we saw with numpy 1?)

test_dask_multiattr_2d - ModuleNotFoundError: No module named 'distributed'

Marking as flaky is fine, but should be unrelated to the No module error, right?

[kounelisagis]

Seems fine, but can you please add some information for future reference about why these fix the errors?

Changed to our own assertion function

Why? (Was this the source of the aborts we saw with numpy 1?)

test_dask_multiattr_2d - ModuleNotFoundError: No module named 'distributed'

Marking as flaky is fine, but should be unrelated to the No module error, right?

Description edited.