GPU: Select fastest gather kernel depending on architecture

[EMCAL-630] Adapt raw fitter spec to changes in raw decoding

- Fix input spec
- Add internal cache for cells for the same IR from multiple links
- Convert internal cache to output vector at the end of the processing
- Add option to select standard or gamma2 raw fitter
- Add abstract function evaluate to CaloRawFitter

[EMCAL-630] Adapt raw fitter spec to changes in raw decoding

- Fix input spec
- Add internal cache for cells for the same IR from multiple links
- Convert internal cache to output vector at the end of the processing
- Add option to select standard or gamma2 raw fitter
- Add abstract function evaluate to CaloRawFitter

Opt. to call carry-over method for last chunk to fix the trailer

DPL: {cpu} expands to always the same CPU

The cpu is guaranteed to have been used by the parent.

DPL: improve error message when matching resources (AliceO2Group#4169)

Changes to allow reading of incomplete/corrupted data

Allows to read data missing TF flags

DPL GUI: display services information in Device Inspector (AliceO2Gro…

…up#4161)

DPL Analysis: Extended histogram registry to use TH2 and TH3 (AliceO2…

…Group#4063)

Fix momentum column (AliceO2Group#4153)