DDSimpleMuonDigi: build the cellID decoder once in initialize()

[TC01]

BEGINRELEASENOTES

• Move the BitFieldCoder used to decode the layer out of the event loop: it is now built once in initialize() and held as a member. Suggested by @jmcarcell in review.
• caloLayout is const

ENDRELEASENOTES

This PR originally replaced the collection-metadata lookup of the cellID encoding with a lookup of the GlobalCalorimeterReadoutID DD4hep constant, so that overlaid beam-induced background in the muon system, whose collections are created at runtime and so carry no encoding in the metadata, could be digitised.

That has been dropped, for two reasons:

• It broke CI on every stack. CLD_o2_v07 does not define GlobalCalorimeterReadoutID (only GlobalTrackerReadoutID), so constantAsString threw and run_DDSimpleMuonDigi failed. Not every detector defines these fallback constants.
• It only fixes one consumer. Every other algorithm reading overlay-created collections has the same problem and would need the same patch.

As a result this PR no longer fixes the overlay digitisation problem — it is now just the cleanup. The intended fix belongs at the producer: OverlayTiming and OverlayTimingRandomMix and will be followed up there.

Given #42 it's possible this shouldn't be merged here; if the desire is for digitizers to live in experiment-specific repositories we could instead make a copy of DDSimpleMuonDigi with these modifications in our fork of k4reco (at least for the time being).

[madbaron]

I think that regardless of where the digitisers get moved to, we can still fix this one centrally here before it gets moved.

[jmcarcell]

I'm checking why it fails for some stacks but not for others

[jmcarcell]

Alternatively the encoding can be set in initialize in the overlay algorithm and then it would be available downstream and in the resulting file; otherwise every algorithm that uses the collections has to be modified. And not every detector defines the fallback cell ID encodings

[madbaron]

Alternatively the encoding can be set in initialize in the overlay algorithm and then it would be available downstream and in the resulting file; otherwise every algorithm that uses the collections has to be modified. And not every detector defines the fallback cell ID encodings

Thanks @jmcarcell, this sounds like a better fix. We'll go make the move from finalise to initialise in the overlay.
This PR now amounts to a bit of cleanup (move decoder out of event loop, drop unused variables, better reporting of error) so it's still worth merging.