Add Lakebase autoscaling support (project/branch parameters) alongside the existing provisioned mode (instance_name) across all Lakebase classes and integrations

Using new autoscaling SDK: https://databricks-sdk-py.readthedocs.io/en/latest/workspace/postgres/postgres.html#databricks.sdk.service.postgres.PostgresAPI.list_endpoints

• Autoscaling resolves hosts via postgres.list_endpoints (READ_WRITE endpoint) and mints tokens via postgres.generate_database_credential, while provisioned continues using database.get_database_instance and database.generate_database_credential
• Strict parameter validation: exactly one mode must be specified — providing neither, partial (project without branch), or both provisioned + autoscaling params raises ValueError
• Engine caching in AsyncDatabricksSession differentiates provisioned vs autoscaling cache keys
  (provisioned::{instance_name} vs autoscaling::{project}::{branch})
• Comprehensive unit tests covering autoscaling host resolution, token minting, routing logic,
  validation edge cases, and engine caching across all four packages
• update other classes using lakebasepool to support autoscaling