sess.span_err instead of 'fail'.  (issue #444)

Hello from SFO Terminal 3!

unify_fn_common had the expected and actual types reversed in one
place. This was causing the type of an occurence of a function f
with type fn(int) -> T to be set to fn(_|_) -> T at a call site like
f(fail); I think this was also making some of the type error messages
come out backwards, but I haven't checked.

Also: ty_bot does not contain pointers

Previously, block_ty returned the type of the terminating
expression of the block (or nil if said expression was absent).
I changed check_expr to write the type of that expression into
the annotation for the block itself, so now block_ty can use the
block's annotation.

The typechecker had a number of special cases for unifying types
with _|_ (as with checking if and alt). But, a value of type _|_
should be usable in any context, as such a value always diverges,
and will never be used by its immediate context. Changed unify
accordingly, removed special cases.

Also make _|_ unify with any type.

Change ty_fn to have a controlflow field. A 'controlflow' is
essentially a bit of data that says whether or not this function
never returns to the caller (if it never returns, that means it calls
"fail" or another "never-returns" function on every control path).

Also add syntax for annotating functions that never return:

fn foo() -> ! {
   fail;
}

for example. Functions marked with ! implicitly have a result type of
ty_bot, which is a new type that this commit also adds.

After all that, give up on "complex" version of type_has_pointers and just manually inspect. Can't calculate property during interning.

Move the ty_item_table into ty::ctxt.items, move variant_info to ty, and start sketching the contains_pointer property on types.

rustc: Unify variables using a single type instead of keeping a list and doing it all at the end. This will give us better error messages.

Except for _task, which is still a keyword.

* Reorganized typestate into several modules.

* Made typestate check that any function with a non-nil return type
  returns a value. For now, the check is a warning and not an error
  (see next item).

* Added a "bot" type (prettyprinted as _|_), for constructs like be, ret, break, cont, and
  fail that don't locally return a value that can be inspected. "bot"
  is distinct from "nil". There is no concrete syntax for _|_, while
  the concrete syntax for the nil type is ().

* Added support to the parser for a ! annotation on functions whose
  result type is _|_. Such a function is required to have either a
  fail or a call to another ! function that is reached in all control
  flow paths. The point of this annotation is to mark functions like
  unimpl() and span_err(), so that an alt with a call to err() in one
  case isn't a false positive for the return-value checker. I haven't
  actually annotated anything with it yet.

* Random bugfixes:

* * Fixed bug in trans::trans_binary that was throwing away the
    cleanups for nested subexpressions of an and or or
    (tests: box-inside-if and box-inside-if2).

** In typeck, unify the expected type arguments of a tag with the
   actual specified arguments.

Still segfaulting on the method-overriding.rs test, though.