Fix nested caplog.filtering() removing filter early

[Fridayai700]

Fixes #14189.

When the same filter is used in nested caplog.filtering() context managers, the inner context's exit calls removeFilter() which removes the filter entirely — even though the outer context still needs it. This happens because logging.Handler.removeFilter() works by identity.

Fix: Before adding a filter, check if it's already present on the handler (from an outer filtering() call). If so, skip both the add and remove, leaving the outer scope's lifecycle in control.

# Before fix: third warning is incorrectly captured
with caplog.filtering(no_capture):
    logging.warning("filtered")       # not captured ✓
    with caplog.filtering(no_capture):
        logging.warning("filtered")   # not captured ✓
    logging.warning("should be filtered")  # captured ✗ (filter was removed)

# After fix: all three warnings inside outer scope are filtered

This is option 1 from the issue — lightweight, no state tracking needed, preserves the existing API contract.