Async reco workflow (after tuning) before vs after GPU decoding

s/TF	original CPU decoding	new GPU decoding
2023 real pp data	4.751 ± 0.038	4.617 ± 0.022
2023 real Pb-Pb data	14.353 ± 0.165	14.151 ± 0.082

Aync reco faster:

By 2.8% for pp
By 1.4% for Pb-Pb

CPU time of o2-gpu-reco-workflow process reduced by (relative to total cpu time):

5.1% for pp
1.6 % for Pb-Pb