feat: add MCP friendly diagnostic APIs

[adarsh0728]

Summary

This PR adds MCP/agent-friendly diagnostic APIs for Numaflow pipelines and mono-vertices. The goal is to expose small, typed, read-only responses that LLMs, automation tools can query directly.

Why

The existing server APIs are useful for the UI, but many of them return raw daemon/Kubernetes shapes, large resource objects, making automated debugging harder because agents first need to discover topology, correlate health/metrics/errors/events across several APIs, and normalize response formats themselves.

What Changed

Added single-purpose typed endpoints for diagnostics:

• Vertex metrics and pending counts.
• Pipeline buffer info and buffer pending counts.
• Pipeline and mono-vertex data health.
• Pipeline watermark lag.
• Pipeline, vertex, and mono-vertex status summaries.
• Vertex and mono-vertex logs.
• Vertex and mono-vertex runtime errors.
• Debug snapshots for both pipeline and monovertex
• Events and Discovery APIs

Backward Compatibility

Existing UI/raw endpoints are preserved. New fields are additive.

[adarsh0728]

@suryapratap-01 let's think from MCP and revamped UI perspective, and take this forward.

[suryapratap-01]

I have already taken your API-first code into consideration and built MCP on top of it. I’ll do another pass to see how we can make it better for the UI as well.