Renato Costa

This PR introduces the first official mixedversion plan mutator: `preserve_downgrade_option_randomizer`. For motivation, see #111595. When this mutator is enabled, the point in which we call `RESET CLUSTER SETTING cluster.preserve_downgrade_option` is...

TFTR! bors r=DarrylWong

No failures on master so far, I think we should be good to backport. blathers backport 23.2 23.1

`TransactionAbortedError` when running workload with `--isolation-level=read_uncommitted --txn-retries=false`: >```Error: error in delivery: ERROR: restart transaction: TransactionRetryWithProtoRefreshError: TransactionAbortedError(ABORT_REASON_PUSHER_ABORTED): "sql txn" meta={id=d757695b key=/Table/111/1/2/1 iso=ReadCommitted pri=0.00665952 epo=0 ts=1708603167.770875286,0 min=1708603167.585376790,0 seq=0} lock=true stat=ABORTED rts=1708603167.770875286,0 wto=false...

This last failure was different: > ```Error: error in newOrder: ERROR: internal error: unexpected error from the vectorized engine: null_value_error: non-nullable column "customer:c_last" with no value! Index scanned was "customer_pkey"...

> Thanks @renatolabs, that's concerning. I'm going to mark as a GA-blocker until we know more. Actually, I suspect there's a good chance this might be part of yesterday's [bad...

TFTR! bors r=srosenberg

Verified that the logic here works by simulating a VM preemption (i.e., running a roachtest and manually deleting one of the VMs on GCE). [0.1 build](https://teamcity.cockroachdb.com/viewLog.html?buildId=14122610&tab=queuedBuildOverviewTab) is also currently in...

TFTR! bors r=srosenberg

Seems like the search algorithm needs to have its parameters updated now that we are using new machine types (#119147) -- cc @srosenberg