Faster decoding of FTRawBankDecoding
Removing sorting and allocating memory for FTLiteClusters.
- Allocating memory: about 15% reduction in CPU time
- Removal of old sorting: about 50% reduction in CPU time
- Adding back of custom sorting for PrFTHits: adds 10% increase in CPU time (about 10 ms, while at least 80 ms can be removed in PrStoreFTHits).
@sesen @rquaglia @sponce @lohenry
Closes LHCBSCIFI-108.
Edited by Marco Cattaneo