Optimise phi-granularity in SiSpacePointsSeedMaker_ATLxk

This MR ports an improvement made by @gavrilen within the scope of ITk reconstruction for use with the ID in run-3.

The seed making step of track reconstruction is run in distinct phi-Z regions, in order to bring combinatorics down to a manageable level and avoid testing unreasonable hit combinations which are not expected to originate from the same charged particle.

The optimal size of these regions in the phi-coordinate depends on the expected trajectory curvature of the lowest-momentum particles we expect to reconstruct, the maximum transverse impact parameter allowed, and the radial locations of the first / last hit that can be part of the seed.

So far, this was chosen very conservatively, using a 400 MeV pt cut-off and the full radial extent of the Si detector. In this MR, the bin size is instead chosen by separately calculating the optimum value for the pixel and strip seeding passes (each uses only one subdetector), and the actual pt cut being run. Then, we pick the more conservative bin size for both passes. This is not as optimised as in the ITk case, where we use two different sets of regions for pixel- and strip-seeding, but since the optimal values in the case of the ID main inside-out pass happen to be very close to another, the performance impact of this simplification should be minimal, while the amount of changes to the code can be kept very limited. In practice, the resulting phi-bin size is reduced by a factor 3 compared to the previously used value.

The impact of this change is a speed-up of the track finding by approximately 20%, and of total track reconstruction on the order of 8%, in mu=60 ttbar events. The efficiencies and fake rates are change only at a negligible level, as the unnecessarily wide phi-regions did not actually add any good track candidates.

For the cast of large-radius-tracking, the regions chosen this way become very wide, slowing down this element of the reconstruction considerably. As the wider regions do not result in an efficiency improvement for LRT, possibly due to other requirements applied in the seeding and track finding, we introduce a flag to disable the new behaviour, which is set for LRT.

Tagging @npetters, @sroe, @gavrilen for information and to give them a chance to suggest changes. Also including @amete as this may (hopefully!) affect the SPOT results.