Improve access to geometry in SciFiTrackForwardingStoreHit

This MR make the following changes in SciFiTrackForwardingStoreHit, allowing for better throughput with exact same result:

• precompute x0 and dx for each mats, allow to remove 5 parameters loads from the cache. +0.5% throughput
• remove the branch for adding the dieGap +1.1% throughput
• set dieGap, sipmPitch, uBegin and halfChannelPitch as constant accross all mats +0.6% throughput

The total provides +2.2% throughput

Reference updates in Moore!440 (merged)

@chasse @sstahl