feat: add progressive and LangGraph BYO kagent examples

[hsteude]

Summary

• add a three-step kagent learning path: basic agent, MCP-enabled researcher, and multi-agent coordinator
• use the faster Go agent runtime for all three declarative examples
• use Anthropic Claude as one replaceable ModelConfig example
• explain the platform-managed gateway-mcp federation and optional custom RemoteMCPServer setup
• scope the researcher explicitly to the discovered fetch tool
• document internal Lab access and external API-key access for each agent
• add a standalone deterministic LangGraph workflow exposed as a kagent BYO agent
• package the BYO agent as a hardened non-root image with an Artifact Registry build workflow

Testing

• linted all YAML manifests and validated the declarative examples against the live cluster CRDs
• observed Accepted for the ModelConfig and Ready for all three Go Agents
• invoked simple-assistant through the internal A2A route
• verified gateway-mcp discovers fetch in the documented single-server setup
• verified web-researcher fetches and summarizes https://prokube.ai with only toolNames: [fetch]
• verified research-coordinator delegates the prokube.ai request to web-researcher and returns its result
• verified the public coordinator Agent Card returns 401 without a key and 200 with its Bearer key
• sent an external A2A message to the Go coordinator through Agent Gateway successfully
• ran 5 BYO workflow unit tests in the built linux/amd64 image
• passed Black, isort, yamllint, image health, A2A Agent Card, and deterministic website-fetch checks
• pushed BYO image tags latest and commit-064d18982baf85544456619742b48a721b295b31 with digest sha256:51354f341a383d747565f27e176ed92775c6274a8773c6b98ed446040086030a

Not Run

• BYO cluster deployment and Anthropic end-to-end invocation because the test cluster was intentionally shut down

Dependency

• stacked on docs: add MCP server examples #58 because agent-with-mcp links to and requires its Fetch MCP example