Speedup of LArRodMonAlg

I noticed that on the ZeroBias stream (that has more raw-data than the main-stream), the LArRodMonAlg is the by far slowest algorithm. Vtune identified the hot-spot: Finding the LArDigit corresponding the the LArRawChannel.

With this change, the execution time changes from:

PerfMonMTSvc                                                      INFO Execute        99        192422.89      0              0                   larRODMonAlg

to

PerfMonMTSvc                                                      INFO Execute        99        28841.04       0              0                   larRODMonAlg

Eg more than factor 6 speed-up.