Changed intensity/parallel values should be respected on retry.

[karol-kokoszka]

If intensity/parallelism values are changed for a running repair, a retry of a failure of that repair should inherit the parallelism/intensity altered settings until that run successfully completes or is started with --no-continue.

As per https://github.com/scylladb/scylla-enterprise/issues/3419

[Michal-Leszczynski]

Do we want change sctool repair control to also permanently change parallel/intensity for the next task runs, or only for the current run?

[karol-kokoszka]

We just want to make sure that the current task run is affected.
https://manager.docs.scylladb.com/stable/sctool/repair.html#reschedule-a-repair
The CLI above can be used to update the task permanently.

[Michal-Leszczynski]

My proposition:

Add intensity and parallel to scylla_manager.repair_run. This way we can prioritize params from DB (when we continue previous run) over standard params (used when running from scratch). This means that params set by repair control will be respected in all runs continuing the run when they were set (e.g. params will be respected when pausing/resuming task).

Side benefit of this solution is that we would be able to display effective parallel and intensity of a already finished task (right now we display those params only for currently running tasks, because we lack this info in DB).