Create a basic test similar to read_write_concern_defaults_application.js/db_reads_while_recovering_all_commands.js to go through all the possible commands and make sure that if a node fails downgrade, the command can run successfully in downgrading, and then when we go back up to upgraded, the command can also run successfully there.

We should test on a standalone node, a replica set, and a sharded cluster

If a command needs to be skipped because it is a new feature, it can be added to a skip list.