Fix bookkeeping of cluster/hypo indices in charged pid algos and downstream. In addition, improves consistency in variable naming (and code comments).
Essentially, only saves CaloCellID (with .all() the unsigned int representation) and not anymore the internal cluster index.
For RelationTables etc, only the SOACollection index is used where appropriate for consistency, as RelationTables only accept SOACollections now (not anymore the inter cluster index).
In addition, some related cleanup / simplifying.
Related MRs: LHCb!3884 (merged), Moore!1901 (merged), DaVinci!802 (merged)
Related issue(s): #357 (closed)