We'd like to make sure that any features that are known not be yet implemented are returning an error like ENOSUP, if this is appropriate for the feature.

One example is the WT_CURSOR->modify operation on layered tables, which apparently was silently a no-op on the follower.  This led to test bugs, since it was expected the feature was implemented and it wasn't.

However, performance related gaps don't need to be included here.  For example, if a cursor marked as bulk load doesn't really do bulk loading in disagg yet, there is no need to return an error, as the functionality (not the speed) of the cursor isn't affected.  Typical functionality tests shouldn't care about this.