Add recurring task to fan out per-connection warehouse loader invocations

[gagantrivedi]

Summary

Register a recurring task in the Flagsmith task processor that, on each tick, iterates over active warehouse connections, logs a metrics row with started_at, and invokes the Lambda loader (#7292) with that run's ID. The Lambda updates the same row on completion via #7293.

Depends on: #7292 (Lambda loader), #7276 (connection model), #7293 (metrics endpoint)

Scope

Recurring task

Register in the warehouse-connections app (same pattern as organisations/tasks.py):

@register_recurring_task(run_every=timedelta(hours=1))
def invoke_warehouse_loaders() -> None:
    ...

Task body

For each active connection (status IN ('connected', 'error'), skipping pending_customer_setup and rotating):

1. Insert a metrics row: { connection_id, started_at: now, completed_at: null, events_delivered: null, batches_dlq: null, error: null } — returns a run_id
2. Invoke the Lambda asynchronously with { "connection_id": "<id>", "run_id": "<run_id>" }
3. Lambda reports completion via Add metrics ingest endpoint for warehouse loader #7293 keyed by run_id, which updates the row

A single failed Lambda invocation must not prevent fan-out to the remaining connections.

Why pre-create the row

• Captures that a run was attempted, even if the Lambda crashes before reporting
• Enables detecting orphaned runs (started but never completed within expected window)
• Add telemetry endpoint for warehouse connections #7289 gets a reliable count of attempts vs completions

Interval

Default run_every=timedelta(hours=1) — matches the PRD's "within 1 hour" latency target.

Acceptance criteria

• Task registered and visible in the task processor's recurring-task list
• On each tick, a metrics row is created and one Lambda invocation per active connection is fired with the run_id
• Connections in non-active statuses are skipped
• Failure invoking Lambda for one connection does not affect others
• Orphaned runs (started with no completion) remain visible in the metrics table