Add score_threshold parameter to similarity search methods

[rodrigotoledo]

Summary

This PR introduces a new score_threshold parameter to improve the quality of similarity search results by filtering out irrelevant matches based on similarity scores.

Changes

• Feature: Added score_threshold parameter to similarity_search, similarity_search_by_vector, and ask methods in the Pgvector adapter
• Feature: Implemented Ruby-side filtering to avoid issues with virtual columns when using score thresholds
• Testing: Added comprehensive test coverage for the new score_threshold functionality and ask method
• Version: Bumped version to 0.1.13 with updated CHANGELOG.md

Technical Details

The score_threshold parameter allows users to set a minimum similarity score (lower distance) to include in search results. When provided, the system:

1. Fetches more candidates than requested (k + 5)
2. Filters results in Ruby based on neighbor_distance <= score_threshold
3. Returns the top k filtered results in the correct order

Usage Example

# Filter out results with similarity score > 0.5 (less similar)
Recipe.similarity_search("chocolate cake", k: 10, score_threshold: 0.5)
Recipe.ask("How to make chocolate cake?", k: 5, score_threshold: 0.3)