Investigate sky subtraction problems

This ticket is a grab-bag of sky-subtraction issues identified by rhl on Slack:

• 115066r4 has over-subtracted sky lines at x=1950, y=1813.
• sky in the pfsMerged files is 0 sometimes
• Horizontal bands in all-spectra plot of 115066.
• 115028r2 has CR at 2250,1130 that's affecting the sky subtraction; increasing the number of spectra in a block (5 --> 20) helps, but should look into the rejection. (fiberId=919 has a bad pfiCenter leading to oversubtraction.)
• There seems to be a difference between the sky flux in the pfsArm and pfsMerged files: the sky-subtracted pfsMerged files aren't zero, but the pfsArm files are. 115029b4, fiberId=2375.
• 115029r2 has some strange light visible to the blue side of some bright OH lines in the sky-subtracted image at 1330,3650.