UMAP methods in the pipeline and worker are different

[alexvpickering]

• The pipeline is using the default uwot (see here)

  scdata <- Seurat::RunUMAP(scdata, reduction = scdata@misc[["active.reduction"]], dims = 1:npcs, verbose = FALSE)

• The worker is using umap-learn (see here)

  data <- RunUMAP(data,
    seed.use = 42,
    reduction = active.reduction,
    dims = 1:pca_nPCs,
    verbose = FALSE,
    min.dist = config$minimumDistance,
    metric = config$distanceMetric,
    umap.method = "umap-learn"

Other differences also e.g. no seed.use specified

NOTE: see here for how to get uwot to not cause segfaults in the worker:

• https://github.com/rexyai/RestRserve/issues/193

tagging @ogibson and @gerbeldo

[alexvpickering]

Seurat::RunUMAP is now only called in the worker.