'spanToNewline' is used to help break apart the source into lines which
can then be partioned into CPP and non-CPP chunks. It is important that
'spanToNewline' not break apart tokens, so it needs to properly handle
things like

  * block comments, possibly nested
  * string literals, possibly multi-line
  * CPP macros, possibly multi-line

String literals in particular were not being properly handled. The fix
is to to fall back in 'Text.Read.lex' to help lex things that are not
comments.

Fixes #837.