document that adata.X is None in v2 docs

[scottgigante-immunai]

Developing new methods is confusing for anyone who has worked with anndata before, as typically you would expect that adata.X would not be None. At the very least this needs to be documented. Partially addresses https://github.com/openproblems-bio/openproblems-v2/issues/120

[rcannood]

The thing is that the AnnData format depends on what's defined in the api/comp_method.yml spec. So in the documentation I can't specify which slots are going to be there other than checking out the contents of this file, reading in the example datataset and printing it out, or building a per-task reference documentation. WDYT?

[scottgigante-immunai]

In that case I would just require that X be set? Seems more straightforward and in line with standard use of AnnData objects. We could consult with scanpy folks if you disagree but it seems pretty reasonable to assert that adata.X is defined.

[rcannood]

In #261 I documented that X is undefined and instead layers are used. This notation is used to make it more explicit what the matrices represent.