shouldRetry() logic in multiInitialSyncApply_noAbort() is over-aggressive

Currently multiInitialSyncApply_noAbort() calls shouldRetry() on all !OK statuses returned from applying any non-command op. If shouldRetry() returns false, the error is ignored. It should probably only do this for error cases that are expecting the shouldRetry() treatment (such as an update not finding a document) rather than for absolutely all errors.