All the CRUD operations (including linearizable read concern) in no_disconnect_on_stepdown.js test, are
taking the failpoints before we acquire RSTL lock. As a result step down does not kill those write operations.

For eg., consider insert operations, hangDuringBatchInsert failpoint got enabled at line at line 354 whereas we acquire the rstl lock only at line 371.

Currently, no_disconnect_on_stepdown test is testing only closing connection part and not killing operations on step down path. So, add new fail points in those CRUD operations such that the fail points are enabled after holding rstl lock and step down kills those operations and operations get an error code as ErrorCodes::InterruptedDueToStepDown.

Also, worth mentioning that currently we already have a test for linearizable read concern which tests killing operations on step down path.