Simplify mapping arguments

[lazappi]

Simplify the mapping arguments in conversion functions to use named vectors instead of nested lists.

Related to (replaces?) #251, fixes #239

• All mapping arguments now expect named vectors instead of lists
• Complex mapping from other slots etc. is now not possible. Users will need to make these changes manually if required.
• The exception is the reduction_mapping argument of from_Seurat()/from_SingleCellExperiment() which allows a more complex list format if users want to specify loadings etc.
• Simple named lists will still work but are not shown in examples
• Unnamed vectors will also work, c("item1", "item2") gets converted to c(item1 = "item1", item2 = "item2"). This is mostly to avoid simple errors and isn't shown in examples.

This helps standardise things so it will be simpler to replace from_Seurat()/from_SingleCellExperiment() with as_AnnData(). More updates to documentation will come in #253.

[rcannood]

LGTM. Minor suggestion

[LouiseDck]

Looks good to me!

[LouiseDck]

What is the reason to use invisible() here instead of just an early return? (purely informative)

[lazappi]

These functions don't return anything, they just modify adata in place. A normal return() returns NULL which might be printed to something, invisible() makes sure the result is hidden. Possibly there is a nicer way to do this.

[LouiseDck]

I think this should be more specific: will it be added as counts or data? In what cases as counts or as data?

[lazappi]

I was trying not to change this here because I thought you would look at it in #255. But I can do it here if that's easier.

[LouiseDck]

No, indeed it will be clarified in #255

[lazappi]

/document

[lazappi]

I'm going to merge this because I think it's holding up other things. I have another PR which should sort out any fixes to the documentation.