Rudder

Since #7524, eventlog migration are parallelized.

Now, the migration process is extremelly sequential, and could accelerated by processing several batch in parallel.

This require to take care of two questions (at least):

- about CPU and I/O load: do we want the migration to use more resource, perhaps impacting more soft real time action like user request
- about on the complexity of the migration process (today, it's dead simple. Here, the migration end is more complexe to assess, because parallel exec, and there is certainly many more things/

The need for that is not very hight, because:

- the migration is a one time event, now async,
- which only impact viewing old eventlogs until the migration ends
- migration file format didn't evolve between 3.0 and 3.2 (and it seems next release).