Further fixes for trigger output handling in reco transforms
This is a follow-up on !27606 (merged) with some further changes:
- Move trigger MT output hacks from reconstruction domain to trigger domain (ATR-20152) because:
- better to have it in one place, not two or more
- better to have it in Trigger domain, not Reconstruction domain
- hacked code can be now reduced as part of the configuration was fixed in !27606 (merged)
- this MR makes it work also for BS->ESD data reconstruction
- Improve EDMDecodingVersion detection by adding the case of reading AOD (like in PhysicsValidation job) - hopefully fixes ATR-20478. Possibly also fixes ATR-20486.
- Work around python memory allocation error when peeking BS file for ROD version (fixes ATR-20480)
cc @tamartin
Edited by Rafal Bielski