Add database constraint for jobs (through partial index)

[tcompa]

We currently have an ednpoint-level check like

    # Check that no other job with the same dataset_id is SUBMITTED
    stm = (
        select(JobV2)
        .where(JobV2.dataset_id == dataset_id)
        .where(JobV2.status == JobStatusType.SUBMITTED)
    )
    res = await db.execute(stm)
    if res.scalars().all():
        raise HTTPException(
            status_code=status.HTTP_422_UNPROCESSABLE_CONTENT,
            detail=(
                f"Dataset {dataset_id} is already in use in submitted job(s)."
            ),
        )

but we observed at least once a situation where two "submitted" jobs co-existed.

---

It'd be best to make this more robust, by also having a postgres unique index on "jobs with dataset_id = X and status = submitted (where X is not null)".
I think a partial index should be appropriate:

• https://docs.sqlalchemy.org/en/20/dialects/postgresql.html#partial-indexes
• "Example 11.3. Setting up a Partial Unique Index" section of https://www.postgresql.org/docs/current/indexes-partial.html

If we are able to implement this at db level, we could:

• leave the current check in-place, to provide a user-friendly error message
• rely on the db check to avoid inconsistent state in some unexpected scenarios (this would result in a 500 response and IntegrityError internal error - which is anyway much better than having two submitted jobs)