Along with this comes the decision to no longer track whitespace and
comments across desugarings in most cases. The main reason for this
decision is that not having to track whitespace simplifies the code and
reasoning necessary.
While implementing desugarings, I also realized that the current comment
model doesn't let me accurately track whitespace across desugarings. I'm
fairly sure I haven't yet found a good model for whitespace and
comments. Once I find one, adding pretty printing support for those
should hopefully be easy though.
I've decided that attempting to create smaller spans is more difficult
and will also lead to worse error messages later on. It makes more sense
to just tag every newly created syntax tree element with the span of the
element it comes from.
