Fix SpatialDimplot(..., interactive = TRUE) to support all SpatialImage types

[dcollins15]

This PR resolves an issue encountered when running SpatialDimplot on SlideSeq or STARmap inputs with interactive = TRUE. Since the functionality is interactive, implementing a proper test is a bit of a PITA so in lieu, the following script can be used to verify that the bug has been resolved:

library(Seurat)
library(SeuratData)

test_case <- LoadData("ssHippo")
test_case <- AddMetaData(
  test_case, 
  metadata = runif(nrow(test_case), min = 1, max = 10),
  col.name = "random_labels"
)

SpatialDimPlot(slide.seq, group.by = "random_labels", interactive = TRUE)

The issue was introduced by yours truly in Seurat 5.1.0 when I added spatial coordinates to the hover output of ISpatialDimPlot. The trouble was that the ScaleFactors generic wasn't implemented for SlideSeq or STARmap, hence the error:

Error in UseMethod(generic = "ScaleFactors", object = object) : 
  no applicable method for 'ScaleFactors' applied to an object of class "c('SlideSeq', 'SpatialImage')"

The fix is to define the generic for the types in question, ScaleFactors.SlideSeq and ScaleFactors.STARmap. Both implementations return a scalefactors S3 class with all its values set to 1. Another alternative is to update ISpatialDimPlot to check the class of object[[image]] and conditionally call ScaleFactors on the appropriate types:

  scale.factor <- 1
  if (inherits(object[[image]], c("VisiumV1", "VisiumV2"))){
    scale.factor <- ScaleFactors(object[[image]])[[image.scale]]
  }

My thinking is that although the scalefactors class is specific to the Visium platform (and therefore kinda "wrong" to populate for other technologies) we're eventually planning on unifying all of the SpatialImage implementations under the FOV class. Until then, it seems like a good idea to take the opportunity to refine the disparate types to expose a consistent interface.

Reported as:

• SpatialDimPlot() with interactive = T doesn't work for slideseq due to ScaleFactors error #9670