-------------------------------------
      Fix the big-tuple-from-desugaring problem
	-------------------------------------

The desugarer generates a tuple from
	- mutually recursive bindings
	- pattern bindings

If either bind a lot of variables, GHC can generate a big
tuple that isn't in the library, with subsequent disaster.

This commit fixes the problem, by using nested tuples.  It
does *not* fix the problem with big tuples written by the
user. And there's still a potential desugarer problem with
parallel list comprehensions that bind a lot of variables
(and parallel array comprehensions) -- but I expect they are
much much rarer.

The fix isn't fully tested yet -- I'll try to do that today.