Mark PythonEdge as "final" per clang-tidy

Classes and methods marked final have some optimization possibilities for the compiler, as some vtable lookups can be eliminated.

I found a consistent but small performance improvement on gt_bench (around 1%).