Fix #13882: Decrement vocab.length when memory_zone clears transient …

[Dzhud]

Fix #13882: Decrement vocab.length when memory_zone clears transient lexemes

Description

This PR fixes issue #13882 where the Vocab.length counter was incremented when adding lexemes but never decremented when memory_zone cleared transient lexemes. This caused len(vocab) to grow continuously even though the actual lexemes were properly removed from the internal hash map, making it unreliable for monitoring memory_zone effectiveness in production environments.

Changes:

• Modified spacy/vocab.pyx: Enhanced _clear_transient_orths() to track and decrement self.length by the number of cleared lexemes, with NULL check for edge cases
• Added comprehensive tests in spacy/tests/vocab_vectors/test_memory_zone.py:
  • test_memory_zone_vocab_length_decremented: Verifies single memory_zone cycle
  • test_memory_zone_multiple_cycles: Verifies multiple cycles
  • Both marked with @pytest.mark.issue(13882)

Testing:
All tests pass successfully:

• Existing memory_zone tests still pass
• New tests verify the fix works for both simple and complex usage patterns
• Tested with reproduction script confirming len(vocab) now correctly decrements
• Code formatted with black and passes flake8 linting

The fix ensures len(vocab) correctly reflects actual lexeme count and matches iteration count over vocab.

Types of change

Bug fix - fixes issue #13882 where vocab.length counter was not properly maintained when memory_zone cleared transient lexemes.

Checklist

• I confirm that I have the right to submit this contribution under the project's MIT license.
• I ran the tests, and all new and existing tests passed.
• My changes don't require a change to the documentation, or if they do, I've added all required information.