Additional schema update

1. Add library_order schema component

uns$library_order = named list, where names = column names in metadata to order (e.g., "cluster_id", "subclass") and values are named character vectors where names are the values of the entries (e.g., "Pvalb", "Sst", "L2/3 IT") and the values are numbers, 0-indexed). Basically the same way hierarchy is embedded, I think.
==> For LATER, these can be used to generated ordered factors for molgen-shiny directory creation. (I still haven't dealt with colors).

Specifically:
library_order["Group"] = ["A" = 0, "B" = 1]; library_order["Subclass"] = ["C" = 0]; etc.

2. Describe child taxonomies ("modes") in documentation.

We need to do this at all. This will be a fairly substantial update to the documentation.

• What are child/subset taxonomies (modes)?
• What is included in them, and where?
• How are they used in scrattch?
• Refine our nomenclature (are do we call them modes, child taxonomies, subsets, etc.)?
• etc.

3. Some adjustments to code/schema to deal with modes.

We need to think a bit carefully about which schema elements need to be assigned to each taxonomy "mode". Some uns elements don't need to be repeated but others do.

• Which don't and which do?
• "default_embedding" should be specified per "mode". We just need a piece in uns to tell us which obsm key goes with which mode (e.g. uns$default_embedding[mode].
• We can place all the mode embeddings into obsm (which doesn't seem to care about the cell dimension even though it should). Currently these are already saved there (I think), but have 0's at the filtered values (the 0's aren't necessarily needed).

@UCDNJJ: For you on Monday if I don't get to it before then. I'll add comments and/or update this issue if I get to things.