Make RNTupleProcessor fully composable

[enirolf]

Explain what you would like to see improved and how.

This issue is meant to collect PRs working towards making the RNTupleProcessor fully composable. Here "composable" means that each type of processor can be create from other processors (e.g., an RNTupleChainProcessor can be created from several RNTupleJoinProcessors or vice-versa). Since this will involve significant changes and additions to the existing RNTupleProcessor code, it seems sensible to split this effort over multiple PRs to hopefully make it a bit easier to review and keep track of the changes.

Three main steps are foreseen:

• Add an RNTupleSingleProcessor for processing single RNTuples, using the same interface as RNTuple[Chain|Join]Processor (PR: [ntuple] Add RNTupleSingleProcessor #17134).
• Refactor RNTupleChainProcessor to use RNTuple[Single|Chain|Join]Processor objects instead of single RNTuples directly (PR: [ntuple] Make RNTupleChainProcessor composable #17393).
• Refactor RNTupleJoinProcessor to use RNTuple[Single|Chain|Join]Processor objects instead of single RNTuples directly.

Additional code and test cleanups will also be tracked here.