Updates and fixes to k = 9 and k = 16 shiny apps

[abspangler13]

• [x] pairwise model boxplots on both apps do not display colors. It's because the test column of sig_genes. Have to remove "BayesSpace" unique(sig_genes$test)
• [x] pairwise results crash at k = 16? make them smaller? filter for fdr < 0.05. Do it in a separate script and save sig_genes object.
• [x] some genes are missing in the k = 9 modeling results. check for these genes in the pseudobulk object, and in the pre and post filtered spe objects. https://neuralplasticity.slack.com/archives/DU3UP2A9W/p1655841909513549. CTGF doesn't exist in the original spe or any of the downstream spe's. PDYN exists in the original spe but was filtered out in the pseudobulking process.
• [x] Update both shiny apps bayesSpace colors so they match those in the paper. Make BayesSpace_colors in colData() https://jhu-genomics.slack.com/archives/C01EA7VDJNT/p1658175550647999
• [x] ask @lcolladotor if there's a better way to filter my k = 16 modeling results so they're small enough to deploy?
• [x] I don't think pairwise stop check in app.R is correct. It doesn't work for my data. Remove *2 because it's before we run sig_genes
• [x] change names so k=9 and k=16 come first
• [x] fix double __in k16 name