************************************************************************* 
There's a problem when two abbreviations overlap, because the number of 
characters saved by using them will not be what was expected.

Right now this is handled, somewhat, by doing a dry run of the 
replacements, noting any that happen fewer than 2 times, and eliminating 
them before doing it for real.

A better solution might be to find all abbreviations that overlap and 
for every set of conflicting ones, test all 2^N-1 possibilities of which 
ones to use and then use the best one.  Ug. 
************************************************************************* 
Add various options for lossy compression:
  - Ignore different whitespace inside replacements
  - Ignore different capitalization 
************************************************************************* 
Add ability to define blacklist on the command line?  Maybe not: undoing 
an abbreviation you don't like is easy to do with sed (at least if you 
haven't also asked for lossy compression). 
*************************************************************************
Replacement character improvements
  - Add more, improve mneumonic algorithm.
  - Try to ensure that the ones that are used are in common fonts.
*************************************************************************
Allow for compressing files that already have UTF-8 characters.
*************************************************************************