api: when a collector in OTHER_COLLECTORS is unreachable, the fetch body does not timeout

[hellais]

This was noticed in "fresh" measurements where the logic is to try to fetch the bodies in this order:

• From disk
• From the list of collectors specified in the OTHER_COLLECTOR setting
• From s3

For some reason if one of the OTHER_COLLECTORS is unreachable, the connection doesn't timeout in time, leading to a failure in retrieving the measurement.

We should look into either setting the urllib3 connection pool timeout or understand how we can make sure the failure is properly triggered so it doesn't stay hanging.

See:

• https://github.com/ooni/backend/blob/master/api/ooniapi/measurements.py#L271