Add Prometheus metrics for TiCDC log output volume (to detect disk-filling log spikes)

[wlwilliamx]

Problem

We’ve hit multiple incidents where TiCDC log output suddenly spiked and filled the disk, causing service impact. Today we have no direct metrics to quantify log volume, so it’s hard to detect/alert early.

Proposal

Expose lightweight Prometheus metrics for log output volume, e.g.:

• ticdc_log_write_bytes_total{output="stdout|stderr|file"}
• ticdc_log_write_lines_total{output="stdout|stderr|file"} (optional)

(Optional) Add write latency / errors later if needed.