add cuda support to mosaicdb

[allen-munsch]

• CUDA Integration & Performance: Introduced comprehensive CUDA support with new Dockerfiles, docker-compose configurations, and benchmarking scripts. This includes performance
  optimizations across the API, hybrid query engine, and indexer to leverage CUDA capabilities.
• Major Core Refactoring (Refactors 0 & 1):
  • Refactor 1: Overhauled query handling by introducing new modules for aggregation, planning, routing, and classification. Replaced the resilience module with a dedicated
    connection_pool and updated DuckDB integration.
  • Refactor 0: Focused on the query engine, introducing helpers, shard router workers, and a worker pool. Significant changes were made to caching, federated queries, indexing
    and search, with the removal of the circuit breaker.
• Advanced Ranking System: Implemented a new, modular ranking framework including fusion, a generic ranker, and various scorers (freshness, PageRank, text match, vector similarit
  to improve search relevance.
• Caching Mechanism Overhaul: Refactored the internal caching system, introducing new ETS and Redis-backed cache implementations, along with dedicated tests and helpers for robus
  cache eviction policies.
• Load Testing & Infrastructure: Integrated k6 for load testing, adding scripts and updating Docker and Nginx configurations to support performance testing.
• Documentation & Development Context: Updated README.md, ARCHITECTURE.md, and added various context/*.txt files to guide ongoing development and architectural understandin
• Minor Improvements: Simplified the embedding service and added a mosaic_demo.sh script for easier application demonstration.