PPRINT-INDENT rounding implementation decisions

[Nikodemus Siivola]

  At present, pprint-indent (and indeed the whole pretty printer)
  more-or-less assumes that it's using a monospace font. That's
  probably not too silly an assumption, but one piece of information
  the current implementation loses is from requests to indent by a
  non-integral amount. As of sbcl-0.8.15.9, the system silently
  truncates the indentation to an integer at the point of request, but
  maybe the non-integral value should be propagated through the
  pprinter and only truncated at output? (So that indenting by 1/2
  then 3/2 would indent by two spaces, not one?)