Nonetheless a few early remarks from my side:

I'd suggest not to use getAsCharData for several reasons:

• outdated and should be removed (CharData was the cell type of the old bufferline impl, those methods only existed for interim transition, but never got cleaned up afterwards because of reasons 😊) - ah right, too convenient in tests 😸
• the method creates an interim array for every tested cell (creating GC pressure for no good reason)

Maybe use cell.getCode() directly instead?

A note on powerline glyphs - most should reside in the Unicode's PUA (private use area), as you correctly test for. But if I remember right there are powerline-like fonts that also use other areas or even single codepoints all over the place (was it nerdfont? idk). Not sure how you want to deal with that, it is a bit like chasing a rabbit, and if you overdo it, you might end up excluding too much ...