When the secondary fetches the oplog from the source, if the source doesn't have anything to provide, we wait for a few seconds before returning data (if we have it) or an empty batch.

Because of this wait, when diagnosing a replication lag issue, it's hard to determine whether the sync source is taking some time to accumulate the data (i.e. if literally scanning the oplog is slow) or whether we're spending time waiting.

We should add a metric that captures the amount of time we spent waiting so that this becomes obvious.