fix: preserve entities without community assignment in _filter_under_community_level (fixes #2348)

[hanhan761]

Summary

Preserves entities without community assignments in _filter_under_community_level instead of silently dropping them.

Issue

Closes #2348

Root Cause

The filter df[df.level <= community_level] evaluates to False for any row where level is NaN. Entities not assigned to any community by the Leiden algorithm have level=NaN after the left join with community membership data. This silently drops all unassigned entities — in some datasets, 93% or more of the entity set — causing CLI queries to return empty or degraded results with no warning.

Fix

Changed _filter_under_community_level to use the pandas-safe expression:

df[(df.level <= community_level) | df.level.isna()]

This preserves entities without community assignments while still applying the level filter to assigned entities. Entities with level=NaN are not "above the requested level" — they simply have no community, and should not be excluded.

Edge Cases Not Covered

• This does not add a warning when a large percentage of entities lack community assignments (the issue author suggested that as an additional improvement). The scope is limited to fixing the data loss.
• The community_level=None path (which skips the filter entirely) is unaffected.

Verification

3 new unit tests:

• NaN rows preserved alongside assigned rows within level
• All-assigned case (regression, unchanged behavior)
• All-NaN case (all preserved)