Automatic exponential formatting in Debug

Automatic exponential formatting in Debug

* {:.PREC?} already had legitimately useful behavior (recursive formatting of structs using
  fixed precision for floats) and I suspect that changes to the output there would be unwelcome.

  (besides, precision introduces sinister edge cases where a number can be rounded up to one
  of the thresholds)

  Thus, the new behavior of Debug is, "dynamically switch to exponential, but only if there's
  no precision."

* This could not be implemented in terms of float_to_decimal_common without repeating the branch
  on precision, so 'float_to_general_debug' is a new function.  The name is '_debug' instead of
  '_common' because the considerations in the previous bullet make this logic pretty specific
  to Debug.

* 'float_to_decimal_common' is now only used by Display, so I inlined the min_precision argument
  and renamed the function accordingly.