[P3][Logs] Aggregate repeated ready, reset, and stale-epoch logs across dispatchers

Parent issue: #4691

Intent

Many dispatcher logs are semantically identical except for the dispatcher ID. Examples include repeated ready, reset, stale-epoch, and handshake mismatch events. This task aggregates those repeated incident shapes so operators can see the scale of the event without reading a line per dispatcher.

Modification points

• Group same-shape dispatcher events by changefeed, event type, and semantic reason.
• Emit summary logs with counts and a small sample of dispatcher IDs instead of one log per dispatcher.
• Preserve individual dispatcher detail only when the event is unusual or explicitly requested.

Entry points

• downstreamadapter/eventcollector/dispatcher_stat.go:521
• downstreamadapter/eventcollector/dispatcher_stat.go:615
• downstreamadapter/eventcollector/dispatcher_stat.go:681
• downstreamadapter/eventcollector/dispatcher_stat.go:688

Done criteria

• Same-shape dispatcher events no longer emit one line per dispatcher by default.
• Summary logs clearly show the affected dispatcher count and representative samples.
• Operators can still pivot to individual dispatcher detail when debugging a specific anomaly.

Read downstreamadapter/eventcollector/dispatcher_stat.go at lines 521, 615, 681, and 688 to trace the current ready, reset, stale-epoch, and handshake-mismatch logging paths. Group events by changefeed, event type, and semantic reason, then verify that default logs show counts and representative dispatcher IDs while unusual or explicitly requested events retain individual detail.