change md_val for TSM, seems to be working good, J501 a bit differetn phi4_sh, rework dec_const, carefull check

Julien's most updated code has all valence derivatives with the sign changed, so md_val needs to flip the sign. Also, the reader does not write "_d1, _d2" in the derivatives anymore for the new data files, so md_val does not look for "_d1, _d2". You need to select by hand which correlators are your valence derivatives, looking at the log file e.g.