Harmonize different types of submitter unknown labels

[sjspielman]

Originally posted by @jashapiro in #553 (comment):

We're finally all debugged and running through with a test file of submitter cell types, which had all of na, NA, and Unknown labels added in for testing.

Just to clarify, in this case, the original file does not actually have all three types? The reason I ask is because I would probably be okay with normalizing "na" and "NA" (and "n/a" and <N/A>) but if this isn't something we are seeing in real data (yet), I would push that off for a while.

---

Down the line, we may wish to harmonize submitter labels in the QC report if/when we encounter a situation when there are multiple types of "unknown" labels. This does not need to be part of the 0.7.0 planned cell typing release (#544)