Create a PBT that tests clustered collection correctness by running queries against regular vs. clustered collections and checking for equivalence. Similarly, backward scans should have the same results as a reverse sort.

We already have the clustered_collection_passthrough suite for the PBTs, but in this suite we compare clustered collections to clustered collections since the clusterAllCollectionsByDefault failpoint is always on.

One way to do this would be to introduce an isClustered flag on the collection model. No need to model the index key, as we only support _id. Then the logic to setup the experiment collection would need to handle this argument, see property_testing_utils.js.