Rationalize download filetypes

[cathoderaydude]

Given recent discussions there are several data issues with Winworld specifically (not adventure) that we should probably try to resolve.

• An enormous amount of downloads (1411 items) are type Archive, but are not actually that type in our new-world filetype definitions - e.g. a ZIP with a single ISO inside should be of type CDISO.
  -- The simplest way to clear out a lot of these is to UPDATE ImageType = CDISO WHERE Name LIKE "%(ISO)%", and then ImageType = 35Floppy WHERE Name LIKE "%(3.5)%") and so on
  -- If that leaves a small number, we can correct by hand
  -- If it leaves a large number, we could crawl the remaining zips on one of the mirrors, read the zip indexes and see how many we can heuristically resolve e.g. if the contents are overwhelmingly of a given type, set to that type

• I believe the architectures x86 and x86-32 were poorly understood by curators in the past. If x86 means 16 bit, then an enormous amount of our stuff could be miscategorized
  -- Current proportions x86 to x86-32 are about 4000:1200
  -- Example apps with x86 tag that I think are not accurate:
  --- Internet Explorer 7/8
  --- VirtualPC 2007
  --- QuarkXpress 4 (verified this one runs on win 10)
  -- Example apps with x86-32 tag that I think are not accurate:
  --- Netscape Navigator 3
  --- I ran out of patience

• Perhaps we could resolve this in large part by doing an UPDATE for all downloads of either type, converting them to x86 or x86-32 based on their age? Anything x86-32 released before the 386 obviously can't be 32 bit.

• x86 should be renamed to x86-16