Move recurring event fetching to rq.cron.CronScheduler #2

Follow-up from review of #1.

Problem

VenFetchJobScheduler in flexmeasures_openadr3/utils/ven_jobs.py keeps its daily ingestion running by re-enqueueing itself at the end of _execute() (in a finally block). This makes the schedule a chain: each run is only scheduled by the previous one.

The finally guards against most in-job failures, but if a worker is restarted (or killed) while handling the job, before it reaches the re-scheduling call, the chain breaks silently and events are never fetched again.

Proposal

Move to rq.cron.CronScheduler, which owns the schedule independently of any single job run. This requires a dedicated rq-cron-scheduler service in docker-compose.yml running something like:

from rq.cron import CronScheduler

cron = CronScheduler(connection=app.redis_connection)
cron.start()

and registering the recurring job with something like:

cron.register(
    self._execute,
    queue_name=app.queues[FETCH_EVENTS_QUEUE_NAME],
    cron_string="0 0 * * *",
    kwargs={"ven_id": ven_client.id, "config_name": sensor_config.name},
)

Note that utc_trigger_time is currently per sensor config, so the cron string would need to be derived per registration rather than hard-coded.

_Originally posted by @Flix6x in https://github.com/FlexMeasures/flexmeasures-openadr3/pull/1#discussion_r3309216061_