Another (s)cPCA example

[LTLA]

This differs from @PhilBoileau's provided Rmarkdown in a few ways:

• I streamlined the entire scPCA discussion by starting off with one of the primed-only batches as the background, which provides an independent dataset compared to all of the other naive/primed cells. This also gets rid of the wacky intestinal worm from the cPCA t-SNE.
• I ended up switching to an arbitrary hard-coded contrasts= due to the above, it seemed to yield a more aggressive removal of the background than relying on the auto-generated clusters=. I suspect this is because the clusters= will also penalizes any removal of the separation between phenotype, which I don't really care about because they're so well-separated anyway.
• Some other reasons for using contrasts= were more practical: it was faster and I didn't want to have to explain to the reader that the general method was still applicable even if you didn't have a priori phenotype information.
• Setting contrasts= to a single value still complains about the need for a non-NULL n_centers.
• scPCA() should handle the coercion of clusters= to an integer internally, as a quality-of-life thing for users.

[PhilBoileau]

This looks great! Thanks for the heavy editing/re-writing. Glad to see that the worm has been removed!

I'll address the issues mentioned in the last two bullets within the next day or so. I must have broken something in the n_centers functionality in one of my more recent updates. Sorry about that!