configure mcp surface vector search and UC functions

[asujithan]

Fixes issue #152

The ucode configure mcp picker only surfaced Databricks SQL, external HTTP connections, Genie spaces, system.ai.* UC MCP services, and Databricks Apps. The other two managed MCP servers documented at https://docs.databricks.com/aws/en/generative-ai/mcp/managed-mcp were not discoverable, so users couldn't add them without hand-editing config:

• Vector Search — /api/2.0/mcp/vector-search/{catalog}/{schema}
• UC Functions — /api/2.0/mcp/functions/{catalog}/{schema}

Changes

• databricks.py: add list_vector_search_catalog_schemas (walks Vector Search endpoints → indexes) and list_uc_functions_catalog_schemas (walk UC catalogs → schemas → probes each for ≥1 user function).
• mcp.py: add VECTOR_SEARCH_SELECTION_PREFIX / UC_FUNCTIONS_SELECTION_PREFIX, vector_search_mcp_servers / uc_functions_mcp_servers transforms, discover_vector_search_mcp_servers / discover_uc_functions_mcp_servers wrappers, plus new branches in build_mcp_picker_choices and _resolve_mcp_selection. New per-(catalog, schema) server names: databricks-vector-search-{catalog}-{schema} anddatabricks-functions-{catalog}-{schema}, slug-normalized with collision suffixes. _discover_mcp_source now wraps each call in a spinner(...) so discovery is no longer a silent wait.
• tests/test_mcp.py: 5 new focused tests covering the transform helpers, picker rendering, and end-to-end registration through configure_mcp_command for both new sources.

Test Plan

• uv run ruff check . clean.
• uv run pytest
• Manual: confirmed both new sources appear oneng-ml-inference.staging.cloud.databricks.com and on e2-dogfood