[TrixNEW]

Closes #21

Adds optional production observability to zphp serve, including:

• Separate admin listener with /health, /ready, and /metrics.
• Prometheus metrics for requests, workers, memory, and uptime.
• JSON server and access logs.
• Request IDs across HTTP/1.1, HTTP/2, static files, and error responses.
• Per-worker metrics with minimal overhead.

All features are opt-in, so default behavior remains unchanged. Added CLI options, documentation, and tests for logging, metrics, concurrency, shutdowns, and edge cases.

Testing

• zig build test passed.
• 197 tests passed on Linux, 196 on Windows.
• Stress tests passed with no observed resource leaks.

Performance

• All features disabled: +0.01% instructions per request.
• Metrics enabled: +0.08%.
• Access logging enabled: +1.25%.
• WordPress benchmarks showed no meaningful regression.

Also found an existing HTTP/1.1 keep-alive performance issue related to TCP_NODELAY. Left it out of this PR to keep changes focused.

[TrixNEW]

Ready for merge