fix: StateManager.Replay on reward actor

[zl03jsj]

in the statemanager.replay method,
after get the msg(specified by 'mcid' param) execution result,
the callback function returns an 'errHaltExecution' to break out the loop in computeTipSetState
but some times computeTipsetstate warps the errHaltExecution with xerrors
that would cause following expression:

(err != errHaltExecution) to be true .

which causes StateManager.Replay returns an error with following message:

unexpected error during execution: callback failed on reward message: halt

[magik6k]

Thanks!