Fix bug on the "clusters (interactive)" tab of run_app() for stitched data.

I used this code to debug things:

```R
library("HDF5Array")

spe <- loadHDF5SummarizedExperiment('/Users/leocollado/Dropbox/Code/_libdcode/visiumStitched_brain/code/05_shiny/spe')
vars <- colnames(colData(spe))

run_app(
    spe,
    sce_layer = NULL,
    modeling_results = NULL,
    sig_genes = NULL,
    title = "visiumStitched_brain",
    spe_discrete_vars = c(
        "ManualAnnotation",
        vars[grep("^precast_k[248]$", vars)],
        "scran_quick_cluster"
    ),
    spe_continuous_vars = c(
        "sum_umi",
        "sum_gene",
        "expr_chrM",
        "expr_chrM_ratio"
    ),
    default_cluster = "precast_k2",
    docs_path = "/Users/leocollado/Dropbox/Code/_libdcode/visiumStitched_brain/code/05_shiny/www",
    is_stitched = TRUE
)
```

Author: lcolladotor

R/app_server.R

@@ -682,8 +682,7 @@ app_server <- function(input, output, session) {
 
         ## Add the reduced dims
         if (reduced_name != "") {
-            red_dims <-
-                reducedDim(spe, reduced_name)[spe$sample_id == sampleid, ]
+            red_dims <- reducedDim(spe_sub, reduced_name)
             colnames(red_dims) <-
                 paste(reduced_name, "dim", seq_len(ncol(red_dims)))
             d <- cbind(d, red_dims)