cl-glut shouldn't error on unknown (special) keys or other enums

[3b]

(or at least not a non-continuable errors)
In particular, windows key (118 and 119) and possibly menu key on windows

also, find and add any missing key defs from current glut implementations.

Not sure which would be better:

• ignore unknown values completely, assuming calling code doesn't know about them anyway
• return the number or intern a symbol named with the number
  this one lets people write key-binding code that (more or less) works with unknown keys, but also lets people write code that will break when a real definition gets added. similarly any saved bindings will break if updated binaries use new definitions.
• ignore unknown values in the normal API, but add a 'raw' API for people to use for bindings?
  probably the best solution, though not sure GLUT is actually worth this much work
• continuable error, so somebody notices they should be added
  annoying if the person who notices is an end-user, but if there are named errors and restarts for specifying what to return, the other options could be implemented as macros and selected by calling code