First class support for zarr input and output

[Zethson]

Description of feature

Judging by the documentation:

nf-core/scdownstream is a bioinformatics pipeline that can be used to process already quantified single-cell RNA-seq data. It takes a samplesheet and H5AD-, SingleCellExperiment/Seurat- or CSV files as input and performs quality control, integration, dimensionality reduction and clustering. It produces an integrated H5AD and SingleCellExperiment file and an extensive QC report.

the pipeline currently only mentions and/or supports h5ad but zarr is becoming increasingly used and supported at scverse. Let's ensure that the pipeline has full support.

[nictru]

The pipeline currently supports several input formats (h5ad, singlecellexperiment, seurat, csv, cellranger h5), then converts all of them to h5ad files. All processes then use h5ad as input and output. In the last step, the final h5ad is also provided to the user as a singlecellexperiment RDS file.

Now the simple way of adding zarr support is adding it as an additional option for the pipeline inputs and providing the final output also as a zarr file.

If 'first class support' for you means being able to rewire all the internals of the pipeline so that it works with either h5ad or zarr that's a larger effort, and also I am not convinced that it's worth the effort given users usually don't see these intermediates

[Zethson]

Yeah, I understand. There's a cost to conversion but it's not a high priority.

Native zarr IO support will at least work better with cloud paths. So let's just focus on this.