Generalize caloriemter hit merging

[ruse-traveler]

Something we've wanted to check in the BHCal DSC is how the response of the detector varies when we group different numbers scintillating tiles together (e.g. 5 tiles grouped into an sPHENIX-style tower).

The CalorimeterHitsMerger algorithm provides roughly the functionality to do this easily, but the catch is that our readout lacks any notion of a sector, tower, or otherwise: it just indexes tiles along eta and phi. Which means we could only merge an entire row in eta or in phi.

However, this algorithm could be generalized in such a way as to allow for the merging of hits according to arbitrary functions. This functionality could be useful in other contexts (e.g. the LFHCal) where calorimeters may want to combine signals during the reconstruction process.

Describe the solution you'd like
One approach could be to utilize similar techniques to what was done for the adjacencyMatrix, peakNeighborhoodMatrix, and various channel-dependent samplingFractions deployed throughout the calorimeter reconstruction chain.

The user could provide a string to the algorithm which tells it how to map the indices of a specified readout field onto desired reference indices. In the above example of the BHCal, there's a simple mapping to get from our individual tiles to a tower of N adjacent tiles in phi:

phi_tower = phi_tile - (phi_tile % N)

Describe alternatives you've considered
Another option could be to just adjust the readout to have the desired fields, rerun the simulation + reconstruction, and repeat as needed. This is, however, extremely labor intensive and lacks any potential synergies with other detectors.