TrigCaloRec: add an option to enforce ROI boundaries for tile cell unpacking

Hello,

As discussed in ATR-23454, even though HLT taus are seeded by L1 TAU ROIs, the tile calorimeter cells that are unpacked and used to build topoclusters can be outside the initial ROI. This MR implements an option to ensure that tile cells are within the eta boundaries of the ROI. It replicates the behaviour we had in Run2 (functionality only used by tau triggers):

• https://gitlab.cern.ch/atlas/athena/-/blob/21.3/Trigger/TrigAlgorithms/TrigCaloRec/src/RoITileCellContMaker.cxx#L131
• https://gitlab.cern.ch/atlas/athena/-/blob/21.3/Trigger/TrigAlgorithms/TrigCaloRec/TrigCaloRec/IAlgToolEFCalo.h#L81-88
• https://gitlab.cern.ch/atlas/athena/-/blob/21.3/Trigger/TrigAlgorithms/TrigCaloRec/python/TrigCaloRecConfig.py#L439

A small difference: compared to the R21 CellInTheRoi method, we only enforce eta boundaries and let the current mechanism handle the phi boundaries. As in R21, the z extension of the ROI is ignored.

No further modification is required as the eta/phi size of the input L1 TAU ROI is now configured upstream (!47406 (merged) ).

This functionality is only enabled for tau triggers, as confirmed by the change in trigger counts.

@damazio , @joheinri , @sutt , could you please check if it looks fine? Tagging @ademaria , @iriu , @carquin for tau triggers. The motivation is to avoid a cluster-level eta cut after topoclustering in the HLT tau code, and rather have the eta cut at the cell level in the HLT Calo code.

Cheers, Bertrand