Add endpoint build logs and server logs retrieval tools

[jralfonsog]

Summary

• Build logs: get_serving_endpoint_build_logs() retrieves container image creation output, dependency installation, and model download steps — essential for debugging failed or stuck deployments
• Server logs: get_serving_endpoint_server_logs() retrieves recent stdout/stderr from the model server, including inference processing and runtime errors
• Auto-resolve served_model_name: Both APIs require a served_model_name param that users rarely know. When omitted, it's auto-resolved from the endpoint's first served entity via get() — zero friction for single-model endpoints

Changes

Layer	File	What changed
Core	serving/endpoints.py	+2 functions + _resolve_served_model_name helper, added ResourceDoesNotExist/NotFound imports
Core	serving/__init__.py	Export new functions
MCP	tools/serving.py	+2 @mcp.tool(timeout=30) wrappers
Tests	tests/unit/test_serving_logs.py	8 unit tests (explicit/auto-resolve, not-found, empty logs)

SDK methods wrapped

Tool	SDK Method	REST Endpoint
get_serving_endpoint_build_logs	build_logs(name, served_model_name)	GET /api/2.0/serving-endpoints/{name}/served-models/{model}/build-logs
get_serving_endpoint_server_logs	logs(name, served_model_name)	GET /api/2.0/serving-endpoints/{name}/served-models/{model}/logs

Test plan

• 8 unit tests pass (explicit model name, auto-resolve, not-found, empty logs, wrong model)
• Ruff lint + format pass (line-length=120, py311)
• Integration tested against aws-fe workspace (ac_demo-demo_data-sklearn_model-serving-endpoint)
• Auto-resolve correctly picks sklearn_model-21 from endpoint config
• Not-found returns clear error message

This pull request was AI-assisted by Isaac