GPU acceleration for COMEBin and SemiBin2

[d4straub]

Description of feature

GPU acceleration seepds up binning with COMEBin and also SemiBin2.
For SemiBin2 that mostly beneficial when no environment (model) is used and that is the case when multiple samples with abundance information (bam files) are given, i.e. when the sequencing depth of multiple samples is used.
COMEBin and SemiBin2 were added without GPU support in #875 and #930, respectively.

There are many nf-core modules with GPU acceleration. Here is a PR that adds gpu labels into the template. And here a docu from a nf-core pipeline about gpu usage. Didnt find anything in the nf-core docu though, maybe missed it.

[jfy133]

The GPUs labels are new, and indeed not well documented... but theoretically you can add the labels to the modules and people hsould be able to start using GPUS (I've never needed to use such a module though... I think further usage may need additional per-user configuration though (e.g. in institutioanl profiles)

[jfy133]

(I think ti's a good idea though!)

[d4straub]

but theoretically you can add the labels to the modules and people hsould be able to start using GPUS

First of, I have no idea. I also thought to myself "add the label and that will work" but that seems not so easy: #875 (comment).

[jolespin]

Would this just be as simple as --engine gpu as a separate option for semibin2 and then building a docker container with cuda support?