[LPD] [Dark] Improve per-memory-cell plots

Description

I noticed that the bad pixels per memory cell plot is kind of useless: the y scale goes up to 1e10, even though the values can only go up to 1. This fixes that, and also improves the offset/noise per cell plots.

How Has This Been Tested?

Running the notebook interactively.

Relevant Documents (optional)

Bad pixels before/after:

Offset & noise before/after:

The Noise plot going down to -10 in the before version is an artifact of how bad pixels were handled. I've changed it to use NaN in calculating these, so the 'Good pixels only' line disappears for cells where all pixels are masked out.

Types of changes

• Bug fix (non-breaking change which fixes an issue)

Checklist:

• My code follows the code style of this project.

Reviewers

@schmidtp @yousefh