added forced unwrapping

[sydow]

This PR implements forced unwrapping and should complete #1859.

We take the approach to parse the postfix operator ! as yet an alternative trailer in atomic expressions, together with ? and indexing, slicing, selection and calls. The operators ! and ? are thus interchangable and expect an expression of optional type as left operand and a sequence of trailers as right operand. It is debatable how useful it is to mix ! and ? in the same sequence, but the following is valid code:

actor main(env):
   d: ?dict[str,list[int]] = { "a": list(range(20)), "c": list(range(30)) }

   a1 = d ?.get("a") ? [1::2] [5] # a1 has type ?int; last !/? operation decides type.
   a2 = d !.get("a") ! [1::2] [5] # a2 has type int
   a3 = d ?.get("a") ! [1::2] [5] # a3 has type int
   a4 = d !.get("a") ? [1::2] [5] # a4 has type ?int
  
   print(a1,a2,a3,a4)
   
   print(d ?.get("b") ? [1::2])  # prints None; failure in ? operation
   print(d !.get("b") ? [1::2])  # prints None; failure in ? operation
   try:
        print(d ?.get("b") ! [1::2])  # raises ValueError; failure in ! operation
        print(d !.get("b") ! [1::2])  # Never executed; would raise ValueError
   except ValueError:
        print("ValueError")
   env.exit(0)

Note that for a successful sequence of operations, the last ?/! operator decides the type of the result. For a failed operation (None encountered in one !/? operation), the failing operation decides the effect: ? gives a None result and ! raises a ValueError.

[sydow]

To reviewers: The changes to LambdaLifter.hs ended up in this PR by mistake; the changes there were done in preparation for an entirely different purpose (removal of classes generation in lambda-lifting). These changes are innocent but should be removed from this PR, but I leave it to run the tests as I must do other things now.

[nordlander]

I support this -- go!

(Assuming the changes to LambdaLifter.hs are moved to some other branch!)