R21.2 DAOD_PHYS: add HLT MET x,y to recover events with saturated L1Calo towers (ATLSUSYBGF-614)

Same as !49873 (merged) but for 21.2 instead of master. This MR is adding HLT MET x,y to the DAOD_PHYS output, for two HLT MET algorithms: cell-based and mht (ATLSUSYBGF-614). This is needed to emulate the HLT MET trigger decision in situations where L1_XE did not fire due to a L1Calo firmware bug affecting data15-16. Explanation given by Jon at the time: "If two towers in opposing quadrants both overflowed then the event wasn't usually written out as the overflows were allowed to cancel each other out. From 2017 this was corrected so that any tower overflow meets the trigger requirement." The motivation for saving both cell-based and mht MET is discussed in ATR-13615. Unfortunately, mht MET is not always calculated, and there's a higher probability that cell-based MET is calculated as it's involved in many combined MET+X triggers.

The size increase is small, it's only 4 extra floats per event. For now, it's added to PHYS only, not PHYSLITE.

Tagging @jburr , @salderwe , @miholzbo , and @calpigia , it would be great to pick this up for the upcoming 21.2 (legacy?) PHYS production. Adding the urgent label for this one.

Cheers, Bertrand