See PM-2524 technical design doc.  This is similar to work done in WT-9910.  Convert test/format to do predictable replay, using/enhancing common support in test_util.  Add an evergreen job for it.  Add temp bug to test predictability as in WT-9910 

As part of this work, add "flush_tier=(enabled)" at configurable intervals to the checkpoint thread.  Whether this gets refactored into calls into testutil to run a checkpoint thread or not, will be determined as part of this effort.