Removed hard-coded dxdy and uses per-sector dxdy.

• Removed dxDy hardcoding from all tracking algorithms in HLT1.
• Test efficiencies in real data.
• Test the UT hits extrapolation and tolerance window cuts.
• Combine UTGeometry and UTBoard into one.

To be merged with: LHCb!4640 (merged), Detector!590 (merged) and Moore!3677 (merged)

Changes to Allen

• dxDy added as a member variable field to UTHits event model and UTGeometry.
• Downstream and matching with UT uses a common UTHitCache.
• UTHitCache now contains per-hit dxDy information. This change neither increase shared memory nor register-per-thread usage in matching with UT algorithm (same can be assumed for downstream).
• UTHitCache encodes and decodes UT hits dxDy and sector dy information with 8 bits which keeps the shared memory usage the same despite adding more information. The encoder-decoder also uses constant memory, hence no increase in register usage.
• UT hits test updated to check dxDy information.
• Forward tracking also uses per-hit dxDy now instead of hardcoded per-layer values.

Testing on Real Data

• The HLT1 thresholds for charm yields are no_ut_tuned_mu4_1000KHz_v1, for both matching and forward sequence.
• The Ks-> pi pi yields uses a custom set of cuts.

Matching with UT Charm Yields

After removing per-sector dxDy, we get 20-25% more charm yields passing through TOS triggers of Hlt1OneTrackMVA | Hlt1TwoTrackMVA. Tested on run 297074.

Forward with UT Charm Yields

The differences here are more drastic since velo-UT part of forward tracking relies heavily on having well-aligned UT detector.

Matching with UT Ks->pi pi

After removing per-sector dxDy, we get 10% more Ks->pi pi in HLT1 (with some selections).