Inconsistent W_ARM value in bias/dark due to outdated redResolution keyword

INSTRM-2200 introduced a robust way of latching the red resolution but left a bug behind

If you move the grating but don't take a shutter exposure, redResolution keyword is not being updated and as a result, some bias/dark can end up a wrong W_ARM value even though the filename is correct.
It's bad because that causes the ingestion to fail in the drpActor.

So I guess that show two problem :

• two different ways to detect the arm, one say low the other say med, that should be unified.
• redResolution has a visit field, that should be checked against the current visit

125108 is one of those visits :

(rubin8-ics) [alefur@shell-ics sps]$ fitsheader PFSA12510812.fits |grep 'W_ARM'
W_ARM   =                    4 / Spectrograph arm 1=b, 2=r, 3=n, 4=medRed