Previously, we were using `size_of` but passing the wrong type in
(`MonochromeSprite` instead of `PathSprite`). This caused us to read
outside of the `sprites` smallvec and triggered the segfault. This
reverts #4078 because I don't think using a `SmallVec` was the issue (it
might have masked this problem though, because we would most of the time
copy from the stack and not from the heap).

With this pull request we are also fixing another potential source of
segfaults, due to checking if we exhausted the instance buffer too late
when drawing underlines.

Release Notes:

- Fixed a crash that could happen during rendering.