Stepdown under ScopeGuard causes MapReduce to terminate when attempting to clean up temporary collections

During MapReduceCommand::errMsgRun.

• We create an ObjScopeGuardImpl0 via ON_BLOCK_EXIT_OBJ on line 1434 here.
• Upon destruction, this guard will run State::dropTempCollections.
• During execution of the block which we have scopeguarded, we step down from Primary:

  [ShardedClusterFixture:job0:shard0:node0] 2018-05-31T14:55:41.662+0000 I COMMAND [conn362] Attempting to step down in response to replSetStepDown command

  [ShardedClusterFixture:job0:shard0:node0] 2018-05-31T14:55:41.671+0000 I REPL [conn362] transition to SECONDARY from PRIMARY

• When we exit the block, the guard is destroyed and invokes SafeExecute, which in turn invokes State::dropTempCollections.
• Because we have stepped down to Secondary, this throws an ExceptionFor<ErrorCodes::NotMaster>.
• Since SafeExecute is not expecting any exceptions to be thrown, it bails out with an std::terminate here.