Part 4 of #98 - 1/f Noise Correction

[JarronL]

Part 4 of #98

Correct 1/f noise striping in stage1:

• Create fnoise_clean.py file that houses code for fitting a 1/f noise model to data and iteratively masks outlier pixels.
• Add cube_fit function, which utilizes polynomial fitting routines from webbpsf_ext to fit linearized unsaturated and unflagged pixel data to quickly generate a kTC bias and slope images.
• Utilize cube_fit to create kTC bias to subtract from linearized ramp data to create clean group-level data
• The slope images from cube_fit are then averaged together and used to produce a high-SNR ramp such that group signal levels match those within a given integration.
• Use 1/f noise filtering algorithm on group level data to model the spatial 1/f striping and subtract it.

1/f cleaning routine adds about 100 msec for a single 320x320 frame. Expect time to increase proportional to frame size. See issue #97 and fnoise.py for more details.

See issue #97 for more information on operations.

Before and after 1/f noise correction for four integrations in PID 1386 Obs 116:

[JarronL]

Already merged into PR #99 in order to resolve merge commits.