The new data handle benchmark from WT-15097 runs a "severe" workload creating many data handles and tracks performance of operations as the number of data handles increase. This should be extended to support disagg storage and layered tables.

Even running this with PALM, we'll get a good idea of how WT scales for disagg. We might need to change the mix of updates vs reads to have a very small proportion of updates so we don't run out of resources for a long run.