bump to v1.1.1

Patch release (no algorithmic changes)

Changes

• Enforced mandatory --model argument with no silent fallback
• Validated scoring model before any embedding inference
• Improved CLI semantic correctness and reproducibility guarantees

Notes

• Alignment algorithm and scoring behavior are unchanged
• This release improves correctness and documentation consistency