Use CNV distributions to identify additional Ewings tumor cells

[sjspielman]

In looking at this I am kind of curious what cell types are in the first hump in the bimodal distribution that we see in the unknown cells are for 822 and 824 when using endo-immune or immune. Using endo + immune together it looks like we might be able to pull out two groups of unknown and could actually label some of the unknown as tumor cells. That's a question for another time, but it might be pretty easy to just plot the total CNV distributions by cell type for that group and see what it looks like.

Originally posted by @allyhawkins in #1144 (review)

---

The goal of this issue is to explore whether we can use inferCNV results for Ewings samples to identify additional tumor cells that were not identified in the cell-type-ewings module.