Optimize rendering performance
Many features have been added since the rendering loop was last profiled and optimized. See where in the code most time is spent and optimize these parts.
LRU caching of create_to_glyph was removed in 4f55947458fe35744de8edffc34fc118fdaecfb6
See alo https://bitbucket.org/pypy/pypy/issues/2365/rinohtype-much-slower-on-pypy3
rst2pdf is much faster at rendering the same Sphinx projects. This might be because it has a simpler styling engine, for example. Anyhow, let's see what rst2pdf is doing (step through it using a debugger) and investigate whether some levels of indirection in rinohtype can be eliminated.