Ensure modularity of PF Reconstruction

[ruse-traveler]

In implementing a baseline PF routine (c.f. #900 ), we should take care to ensure the modularity of the overall routine. The WIP that was started in #1186 risked becoming monolithic, but could be reworked to be modular.

A possible approach:
This was discussed during the July 2024 UGM e-Finder/Holistic Reco Workfest (see the summary talk here), and there it was suggested that we can conceptualize PF in 3 stages:

1. Connection: where associated tracks, clusters, and PID hypotheses are identified and bundled together;
2. Arbitration: where these associated objects are reworked and distilled into particle candidates; and
3. Regression: where the particle candidates are converted into reconstructed particles.

[ruse-traveler]

Associated PRs:

1. Connection: could extend work from Add Track-Based Cluster Merging Algorithm #1406 to partially address this.
2. Arbitration:
   • track-cluster subtraction: Implement Track-Cluster Subtraction (PFA1) #1627
3. Regression: TBD