Add "blocked/not blocked" in job count metrics

[severo]

Now that we block datasets, the job count metrics are a bit misleading, because they still include the jobs of blocked datasets. We need to be able to filter them out, because they are outside of the queue during the blockage.

[severo]

It's very important because it commands the auto-scaler. We don't want the number of pods to remain high while all the jobs are blocked.

[severo]

reference: #2279 (comment)

[AndreaFrancis]

Pending (#2949 (comment)):

• Decrease jobs when a dataset is blocked
• Find a way to include the jobs in the counter once the dataset has been released from blocked list

[severo]

Find a way to include the jobs in the counter once the dataset has been released from blocked list

I think it's OK to wait for the next cron job (every 10 minutes in prod, compared to the duration of the dataset blockage: 6 hours)

[severo]

We still need to be able to filter out the blocked jobs from the charts in https://grafana.huggingface.tech/d/i7gwsO5Vz/global-view?orgId=1 (or show two curves: blocked / not blocked) because it's misleading otherwise