Investigate pExpr in the Markdown printer

[BilalM04]

Current state of pExpr is detailed in #3823 (comment).

pExpr re-defines all the patterns that are different, then delegates to the TeX function in a catch all statement. But since the pExpr function involves recursive calls, the patterns which make a recursive call are redefined as well, leading to some duplication. This is because the current TeX.pExpr does not account for variabilities in the LaTeX commands.

The pExpr function in the Markdown printer needs to be revisited to avoid duplication. Additionally, the TeX.pExpr function needs to be revisited to account for variabilities in syntax.

[JacquesCarette]

I think the (inevitable) conclusion here is that pExpr implements a context-dependent translation process. So we need to understand what the variabilities are, and what drives the choices between the options. Also, because of the clear duplication, it may well be that we need to do a bit more abstraction to get there.

One technical point to keep in mind: right now, we generate a raw Doc. I think we're hitting the limits of this context-free method of doing things. We might need to do

newtype TeXDoc = TeX Doc
newtype MarkdownDoc = MD Doc

which are then still essentially a Doc but we have additional information available that we can dispatch on (with Haskell's type classes).