[Draft]added gcp support in llm engine

[muhammad-zoaib-scale]

Summary
Add GCP (Google Cloud Platform) support to model-engine, enabling deployment on GCP infrastructure alongside existing AWS and Azure support.
Changes
New GCP Gateway Implementations:
gcs_filesystem_gateway.py - GCS file operations using smart_open library
gcs_llm_artifact_gateway.py - LLM model artifact storage and retrieval from GCS
gcs_file_storage_gateway.py - High-level file storage interface for GCS
New GCP Repository Implementations:
gcs_file_llm_fine_tune_repository.py - Store/retrieve fine-tune job templates from GCS
gcs_file_llm_fine_tune_events_repository.py - Store/retrieve fine-tune events from GCS
New Queue Delegate:
redis_queue_endpoint_resource_delegate.py - Redis-based queue lifecycle management for GCP (uses Memorystore)
Modified Files:
api/dependencies.py - Added GCP conditionals for dependency injection
celery_autoscaler.py - Added GCP_MEMORYSTORE_REDIS_BROKER constant
infra/gateways/init.py - Export new GCS gateway classes
infra/repositories/init.py - Export new GCS repository classes
requirements.in - Added google-auth and google-cloud-storage dependencies
Architecture Decisions
Celery Broker: GCP uses Redis (Memorystore) - same as ElastiCache pattern
Celery Backend: Uses Redis for task results (not GCS, avoiding need for custom Celery backend)
Storage: Uses smart_open library with GCS client, consistent with S3/ABS patterns
Queue Management: Redis queues are implicit (created on message push), so delegate handles lifecycle tracking and metrics