Run turbulent heat flux calculation after the merge of l3 files to save time

[BaptisteVandecrux]

At the moment, the same calculation (and subsequent daily and monthly averages) are being done on old transmissions, that sometimes have now been replaced (in the l3 product) by data from raw files. This is redundant and maybe not necessary, especially because we do not need those files for the construction of the time-critical bufr files.