fix: don't drop single cell matrix

[maxkarlsson]

This is a fix of an issue that arises when an assay is subset to a single cell. For example, calling for a data layer in such a subsetted would return a numeric vector instead of a Matrix:

> seur <- LayerData(subset(pbmc_small, cells = 1))
> class(seur)
[1] "numeric"

> seur <- LayerData(subset(pbmc_small, cells = 1:2))
> class(seur)
[1] "dgCMatrix"
attr(,"package")
[1] "Matrix"

This fix makes sure that the returned object is a Matrix, and I have added a test to control for that. This may seem like a specialized case, but it is an issue that creates trouble for certain deep sequenced assays where inspecting individual cells are more common.