feat(store-postgres): Add PostgresStore for LangGraph.js

[abogacki]

Implement PostgreSQL Store with Vector Search Capabilities

Overview

This PR implements a PostgreSQL-based store for LangGraph.js inspired by #887 and Python implementation. The implementation includes comprehensive vector similarity search, hybrid search (combining vector and text search), and TTL management for automatic data expiration.

Key Features

• Modular architecture with specialized components for different responsibilities
• Vector similarity search with support for multiple distance metrics (cosine, L2, inner product)
• Hybrid search combining vector similarity and full-text search
• Advanced filtering with rich query operators ($eq, $gt, $lt, $in, etc.)
• Automatic TTL management with configurable expiration settings
• Support for PostgreSQL's pgvector extension for efficient vector operations

Implementation Details

• Split the monolithic implementation into specialized modules:
  • database-core.js: Core connection and transaction management
  • database-setup.js: Schema initialization and migration
  • crud-operations.js: Basic CRUD operations
  • search-operations.js: Advanced search capabilities
  • vector-operations.js: Vector embedding and similarity calculations
  • ttl-manager.js: Expiration and cleanup management
  • query-builder.js: SQL query construction
• Added comprehensive TypeScript types for better developer experience
• Implemented efficient text extraction from JSON documents using configurable field paths
• Added extensive test coverage for all features

[abogacki]

I figured that this won't have any effect on other integration tests relying on this docker-compose file, and it will enable store-postgres to be tested against vector search. I've also added docker-compose.test.yml file in /libs/store-postgres

[abogacki]

I've decided to group and split integration tests by tested module/functionality, as originally the file had over 1,5 k lines of code and would be harder to maintain

[abogacki]

These types/objects are currently included in @langchain/langgraph-checkpoint library, so I've added it as a dependency. I was kind of torn apart between copying them, or creating separate module for @langchain/langgraph-store (similarily to @langchain/langgraph-checkpoint). Lmk which approach would you like me to go with (or include the changes yourself 🙂)