Update external instructions celltype reference information

[sjspielman]

After #502, we left some TODOs in the external-instructions.md:

scpca-nf/external-instructions.md

Lines 361 to 377 in 74b744b

	#### `SingleR` references
	By default, `SingleR` annotation uses references from the [`celldex` package](https://bioconductor.org/packages/release/data/experiment/html/celldex.html).
	TODO: AVAILABLE REFERENCES ARE LISTED HERE.
	Please consult the [`celldex` documentation](https://bioconductor.org/packages/release/data/experiment/vignettes/celldex/inst/doc/userguide.html) to determine which of these references, if any, is most suitable for your dataset.
	`SingleR` reference files are formatted as `<singler_ref_name>_celldex_<celldex_version>_model.rds`.
	For example, consider the reference file `BlueprintEncodeData_celldex_1-10-1_model.rds`:
	- The reference name is `BlueprintEncodeData`.
	- The `celldex` version is `1.10.1` (for the file name we substitute dashes for periods).
	#### `CellAssign` references
	By default, `CellAssign` annotation uses marker gene set references from [PanglaoDB](https://panglaodb.se/), as compiled by the Data Lab to represent common organ/tissue groupings.
	TODO: AVAILABLE REFERENCES ARE LISTED HERE.
	The specific organs used to compile marker gene references are listed in [`celltype-reference-metadata.tsv`](references/celltype-reference-metadata.tsv).

Currently #565 is establishing an approach to be able to quickly see all available models. We should update the instructions to point external users to those spots.

[jashapiro]

Noting that we can link to https://scpca-references.s3.amazonaws.com/celltype/singler_models/singler_models.tsv and https://scpca-references.s3.amazonaws.com/celltype/cellassign_references/cellassign_references.tsv for people to look up currently recommended cell type references.

Note that these files will not include older versions which would still exist on S3, since we do not plan to delete those.

If we want to get really fancy, we could write an action to pull these down, but I think a link is probably sufficient.

[jashapiro]

closed by #573