Basically this is like lint/gofmt for md docs.

It basically does the following:

• iterate over all files in the given doc root.
• for each file split it into a slice (mungeLines) of lines (mungeLine)
• a mungeline has metadata about each line typically determined by a 'fast' regex.
  • metadata contains things like 'is inside a preformmatted block'
  • contains a markdown header
  • has a link to another file
  • etc..
  • if you have a really slow regex with a lot of backtracking you might want to write a fast one to limit how often you run the slow one.
• each munger is then called in turn
  • they are given the mungeLines
  • they create an entirely new set of mungeLines with their modifications
  • the new set is returned
• the new set is then fed into the next munger.
• in the end we might commit the end mungeLines to the file or not (--verify)