Bugs Causing Faulty Wavelength Solution for the RM Grating

[yuguangchen1]

There are two bugs associated with the FindCenter primitive that cause the wavelength solution to fail for the RM gratings.

1. When solving the dispersion, the reference arc spectrum is repeatedly modified by the Tukey window, so that the useful section of the spectrum progressively narrows.
2. The correct keyword in the kcwi.cfg file to modify the Tukey alpha value is TUKEYALPHA. However, TAPERFRAC is used in the current version. This causes the custom `alpha=0.7' configuration for the red channel to not register in the actual reduction.

Solutions to both problems are provided in: 99e05af

[MNBrod]

Thanks for this! I'll test both of these out and implement them into the main pipeline branch once I verify their behavior. I'm not 100% convinced the first one is an issue for most users (when running in parallel) but it certainly seems possible. The second one is just clearly a bug, thanks for catching that!