[ATR-21112] Input maker refactor - custom ROI Tools for EventViewCreator

As per the JIRA, the main changes here are

• Supply the correct ROI to be executed via a new Tool to the EventViewCreatorAlgorithm
• Change InputMakerOutputDecisions from an array of output handles to a single output handle. With per-Leg DecisionIDs for combined chains, we always merge the inputs from the N Filter algs back to a single output.
• Ability to merge based on Feature rather than based on ROI. This is future planning as it will be needed for the muon fullscan case (all FS muons have a common FS ROI). Could also be used for jets.
• Ground work to enable re-use of View instances from previous Steps (a.k.a. "caching") where two EventViewCreatorAlgorithm in different Steps share the same reconstruction sequence. This won't actually be activated until more groundwork has been put in place in the EDM export.

Private tests reveal some configuration issues which this refactor may have brought to light, this MR will need some iteration from its initial point.

A follow up MR will go the extra step of moving the extra functionality of EventViewCreatorAlgorithmWithMuon, EventViewCreatorAlgorithmWithJet into EventViewCreatorAlgorithm - a mix of ROI tool and some new templated helper functions.

Update March 11th

• Re-added the ability to link multiple parent Views. Used e.g. for Egamma PrecisionCalo stage (accepts Decision Objects which have been processed in parallel by both the Elelctron and Photon paths and hence have two View links). Similar for tau with two flavors of calo.
• Temporarily removed BPhys as their BPhysics Decision Objects with two initialRoI are now explicitly incompatible with the Muon Step3 input maker. These need re-working (ATR-21092)
• q221 now runs to completion, investigating count changes.