Pretty-printer megaticket: fix bugs which produce weird or unparseable code

[pchiusano]

Collecting various pretty-printing bugs here. Feel free to leave a comment with other examples, or with a refinement of any of the existing examples.

• As noted in Uncontroversial fix to rendering of tables in docs #1897 lists that span multiple lines get weird treatment. See doc.output.md for an example. I think it still parses fine but it looks pretty weird.
• Revisit breaking of doc blocks. Again look at doc.output.md.
• f x y should render as f x y or

f 
  x 
  y

But I think I've seen it render as:

f
x
y

which will not even parse. I don't have a minimal example of this one though, and this specific example works as expected I think.

[hojberg]

Here's a few more:

• Preserve record syntax in view and edit #2260
• Pretty printing multiple list head matches removes parens #2224
• Case of wrong indentation of |> in printed code #2165
• edit + slurp roundtrip fails for long lines with repeated operators #1035

[runarorama]

Some issues that collide with this:

• Almost no Doc.Deprecated documentation in Base can be parsed correctly after pretty-printing. Rather than hack this in the pretty-printer, we should just convert all docs to the new type. Tracked in Convert from Doc.Deprecated to Doc base#39
• A number of terms in Base have an overly general ability type, which interacts with general ability inference goofiness and causes a need for additional type annotations: Ability inference goofiness ticket #1271
• Those terms sometimes can't be updated because the inferred type matches the new type signature. There should be a way to replace the term and/or type signature of a hash #2276