V23 compiled figures - Githubissues

Closes #1676

This PR re-compiles figures that needed recompiling. I found a few other sneaky axes to update "sample" -> "tumor" that were previously missed. There are now PDFs and PNGs of new compilations. Here's the rundown (hopefully roughly complete...) of why each of these compilations changed, figure-by-figure:

Figure 1
- Panel B y-axis is updated to say "tumors"
- Panel B now reads "Ependymoma" and not "Ependymal tumor"
Figure 2
- All oncoprints changed with V23 independent samples
- Embryonal legend was reordered
Figure 3
- Panels A/B changed with V23 independent samples
- The panel A y-axis needed a "sample" -> "tumor" update, so I updated the interaction-plots code, regenerated figures, and copied over the new figure panel to figures/pdfs/.
- Panel C regression changed with V23 independent samples
- Panel D changed with V23 independent samples
- Panel E samples -> tumor change
Figure 4
- Panels G and H (2 of the survival models) had slight changes in V23
Figure 5
- UMAP legend now reads "Ependymoma" and not "Ependymal tumor"
- I replaced the GSEA heatmap with new V23 version, since it's a heatmap and who really knows about diffs.
- Panels C and E sample -> tumor change
Figure S4
- Panel A sample -> tumor
Figure S6
- EPN panel title now reads "Ependymoma" and not "Ependymal tumor"
- Panel E and F sample -> tumor change
- Note that Panel F was re-generated in #1682 but I missed adding the new PDF! It's added here.

Worth noting that other compilations were updated previously:

Figure S3: https://github.com/AlexsLemonade/OpenPBTA-analysis/pull/1681
Figure S5: https://github.com/AlexsLemonade/OpenPBTA-analysis/pull/1669

Also worth noting that this PR does not do the tumor purity supp figure described in #1665. That will happen in its own PR.

Reviewers please let me know if/where I can clarify inventory here!

AlexsLemonade / OpenPBTA-analysis

V23 compiled figures #1683