WHAT

Since we've decided and tested the num.recovery.threads.per.data.dir value in MGDSTRM-9019, we should set the value to kafka instance.

WHY

To improve the return to service time following abnormal broker shutdown

HOW

It is best if systems adapt to their environment, rather than exposing another knob for a human to turn/forget to turn.

Can we make use of cpu fleetshard passes to the broker container with appropriate max/min clamps so it can sensibly adapt to different cpu resource limits?

I note that this config is per data dir, so for future proofing, I suggest putting the number of data dirs into the equation too..

DONE

Set the value in fleetshard.