feat(server): refactor server ingestion to sink

[namrataghadi-galileo]

Summary

Refactored the server observability write path to use the shared ControlEventSink abstraction while preserving the existing /api/v1/observability/events behavior and Postgres-backed OSS storage.

Added a default OSS server sink that adapts the existing EventStore write path into the shared sink contract.

Kept the current DirectEventIngestor usage stable by allowing existing store-based construction to continue working, while routing writes through sink semantics internally.

Scope

User-facing / API changes

• No intended user-facing or HTTP API changes.
• /api/v1/observability/events request/response behavior remains unchanged.
• IngestResult and API response accounting semantics are preserved.

Internal changes

• Updated the shared sink contract to support both sync and async sink writes.
• Added a server-side default sink backed by the existing EventStore/Postgres path.
• Refactored DirectEventIngestor to write through ControlEventSink internally.
• Preserved existing server wiring by wrapping EventStore inputs into the default sink internally.
• Added test coverage proving the ingestor can accept a sink directly.

Out of scope

• Config-driven sink selection.
• Alternate server sinks such as ClickHouse, OTEL, Kafka, or vendor-specific sinks.
• Changes to server query/stats read-path behavior.
• Changes to SDK behavior beyond the minimal shared sink contract compatibility needed for server support.

Risk and Rollout

Risk level: Medium

Rollback plan:

1. Revert the shared sink contract async compatibility change.
2. Remove the new server sink adapter.
3. Restore DirectEventIngestor to writing directly to EventStore.store(...).
4. Keep the server endpoint and startup wiring unchanged.

Testing

• Added or updated automated tests
• Ran make check (or explained why not)

Validation has not been run by me in this branch flow; recommended focused server tests should be run locally/CI.

• Manually verified behavior

Checklist

• Linked issue/spec (if applicable)
• Updated docs/examples for user-facing changes

No docs/examples updates were needed because this story is internal-only and preserves existing API behavior.

• Included any required follow-up tasks

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[lan17]

I don't see a new code issue in the current server-ingest diff, but this PR is still stacked on the pre-update head of #164 and GitHub currently shows it as non-mergeable. Please restack/rebase onto the current PR 164 head and I'll do a final pass on the resolved diff.

[lan17]

I don't think this refactor pays off if ControlEventSink.write_events() can now return either SinkResult or Awaitable[SinkResult]. That makes the core contract harder to reason about: the server has to normalize with resolve_sink_result(), while the SDK has to guard and fail at runtime if someone wires in an async sink. I'd keep the sync and async contracts separate and adapt once at the boundary instead of widening the shared interface itself.