Allocate additional pixel mask bits to NIR data.

The NIR detectors have many interesting pixels, and I propose that it would be good to persist their categorization in more detail. In this ticket, adding a few more details about BAD pixels. The idea is to always mark these pixels BAD, but also set some other bit to help track provenance and evolution.

Most BAD pixels come from either the defects product (generated from the master darks, and mostly but not entirely "hot pixels") or the linearization product (generated during linearization curve fitting, and mostly but not entirely "dead pixels"). Between the two of them, maybe 200-300k pixels.

I thought we had a DEFECT mask which simply showed the content of the defects list and so named the plane for those pixels "DEFECT" in PIPE2D-1843; it turns out that may not be the case, so perhaps a different name would be better.

BAD pixels which come from the linearization fitting I have been putting in a LINEARITY_DEFECT plane. Most are simply dead or only faintly responsive, but there are many far more exotic behaviors as well.

[ Many BAD pixels are in both collections. I need to be more careful about how that is handled (linearization passes though a bad pixel plane which is actually just a defects product mask, so ISR can actually use either, leading to a potential provenance botch) ]

The third category is for the pixels which are picked up in ISR, nearly all because their rates are not stable across the ramp or because they are have too many RTS or other (often uncorrectable) glitches to handle. A few thousand per 900s ramp, so not a huge number. I have been putting these in an "UNSTABLE" plane. I imagine that once we start recording these many will be seen as always BAD and can perhaps be put in a proper DEFECT list.

There is an existing "SUSPECT: the pixel is suspected to be non-linear" plane. I thought that could be used for UNSTABLE, but I think the usage is to mark iffy pixels which are not actually BAD.

Marking just the source of each defect does not sound right, but I wanted to avoid identifying and populating a full bestiary. And labeling just the 250k cataloged BAD with some other bit lets us at least see the few thousand others we might want to pay more attention to.

Given that there is no official "DEFECT" plane and I don't need to bend to it, I will propose the following. Add:

• DARK_DEFECT: the pixel was categorized as BAD from darks
• LINEARITY_DEFECT: the pixel was categorized as BAD from linearization flats.
• UNSTABLE: the pixel was labeled BAD during ISR.
• and maybe later UNSTABLE_DEFECT: the pixel was categorized as BAD from UNSTABLE history.

price? rhl?