Right now, ReplicatedFastCountManager::shutdown() is called during step down as part of the `onStepDownHook()`. This flush silently fails with the error

[js_test:wt_size_storer_cleanup_replica_set] d20040| {"t":{"$date":"2026-03-02T16:26:52.960-05:00"},"s":"W",  "c":"STORAGE",  "id":7397500, "ctx":"conn2","msg":"Failed to persist collection sizeCount metadata","attr":{"error":{"code":10107,"codeName":"NotWritablePrimary",

Fix this error by (1) upgrading the warning to at least an error or an assertion and (2) changing where we call shutdown(), perhaps to an onBeforeStepDownHook() function.

As part of this change, update tests that should be compatible which contain the TODO for this ticket number.

EDIT: This problem is not solvable by flushing before stepdown. If step down is initiated, even if we flush our changes to the oplog, it is very unlikely those changes get replicated because we are transitioning to secondary. Since this issue goes away after recovery is implemented, we will instead exclude the failing tests (for now).