Fix flaky filter test

[phoebusm]

Reference Issues/PRs

What does this implement or fix?

Use initialized numpy array for test_filter_unfilterable_data instead of unintialized one in the hope that the flakiness will be gone, as uninitialized value could result in np.nan which are considered not equal in np.allcose

In [1]: import numpy as np

In [2]: arr = np.full((3,4), np.nan)

In [3]: arr2 = np.full((3,4), np.nan)

In [4]: np.allclose(arr, arr2)
Out[4]: False

In [5]: arr = np.zeros((3,4))

In [6]: arr2 = np.zeros((3,4))

In [7]: np.allclose(arr, arr2)
Out[7]: True

Any other comments?

Checklist

Checklist for code changes...

• Have you updated the relevant docstrings, documentation and copyright notice?
• Is this contribution tested against all ArcticDB's features?
• Do all exceptions introduced raise appropriate error messages?
• Are API changes highlighted in the PR description?
• Is the PR labelled as enhancement or bug so it appears in autogenerated release notes?