Add Siphon internal events table

What does this MR do and why?

This MR creates the siphon_internal_events table to hold replication Siphon statistic in ClickHouse. With this table we can periodically calculate the Siphon data lag per table:

 SELECT
      postgresql_schema,
      postgresql_table,
      argMax(dateDiff('millisecond', produced_at, consumed_at), consumed_at) AS data_lag_ms
  FROM siphon_internal_events
  WHERE event_type = 1
    AND postgresql_schema <> ''
    AND postgresql_table <> ''
  GROUP BY postgresql_schema, postgresql_table

   ┌─postgresql_schema─┬─postgresql_table──────┬─data_lag_ms─┐
1. │ public            │ issues                │         493 │
2. │ public            │ merge_request_metrics │         256 │
3. │ public            │ projects              │         979 │
4. │ public            │ merge_requests        │         520 │
   └───────────────────┴───────────────────────┴─────────────┘

Why is this important?

Querying data from a different data stores has different consistency and timing guarantees. By knowing the data lag by table, we could do the following:

• Application may decide to use a different data source (PostgreSQL) in case the CH data lag is high.
• Provide lag data popover on our visualizations, explaining that the data is not 100% up to date. (For example "red" exclamation mark when the lag goes above 1h).

Schema

column	description
uuid	Unique identifier for the event, generated by the consumer when the metadata is recorded.
stream_identifier	The consumer stream (streams[].identifier) the package was consumed from.
producer_application_identifier	application_identifier of the producer that published the package.
consumer_application_identifier	application_identifier of the consumer that processed the package.
postgresql_schema	Source PostgreSQL schema of the replicated table.
postgresql_table	Source PostgreSQL table.
timestamp	When the consumer recorded the event (generated locally, UTC). Also the partition/TTL key.
produced_at	When the producer published the package to NATS (from the siphon_stream_created_at header). Falls back to consumed_at when the header is absent (e.g. the snapshot-finished marker).
consumed_at	When the consumer finished processing the package (same instant used for the siphon_data_lag_ms metric). consumed_at - produced_at is the end-to-end lag.
package_events_count	Number of row-level events in the package (len(events)).
event_type	Package classification; see the event_type values table above.
version	ReplacingMergeTree version column; defaults to now64(6) on insert.
deleted	ReplacingMergeTree soft-delete flag; defaults to false. Present for engine consistency; the consumer only inserts.

event_type values:

value	meaning
1	Standard CDC package event
2	Initial snapshot data (first event carries the snapshot operation)
3	Snapshot finished
4	ClickHouse refresh event

References

In-product observability about replication stat... (gitlab-org/analytics-section/siphon#177 - closed)

MR acceptance checklist

Evaluate this MR against the MR acceptance checklist. It helps you analyze changes to reduce risks in quality, performance, reliability, security, and maintainability.