Also, some other changes that came up along the way:
- add a 'blk' region for the current block.
- detect unused type/region variables.

This alignment is the "preferred" alignment of a type, which is not
necessarily the alignment the compiler will use when packing the
type into structures - that is the "ABI" alignment (in LLVM terms).

On x86, 64-bit ints have 8-byte preferred alignment, but 4-byte
ABI alignment.

- paths can now take region parameters, replacing the dirty hack
  I was doing before of abusing vstores.  vstores are now a bit
  of a hack though.

- fix various small bugs:
  - we never checked that iface types were compatible when casting
    to an iface with `as`
  - we allowed nonsense like int<int>
  - and more! (actually that may be it)

This helps with prototype inference and avoids some 'must be known
in this context' errors.

Issue #2111

This reverts commit 6e909e38.

This reverts commit 06afe21c.

I had to xfail one existing test case (class-implements-int) because,
I think, of the same bug described in #2272.

    enter_opt was handling the (*) case wrong and causing a bounds check
    failure. Fixed it (the test case is one of the extracted ones from
    the reference manual)

Closes #2220

Test case disabled until a memory-leak issue is resolved.

Closes #2255

Closes #2185

In the "not all paths return" error message, typestate was printing
the AST type from the fn decl, not the ty::t type. This ended in tears
when the AST return type was "ty_infer". Now it looks up the function node
ID's type and uses util::ppaux::ty_to_str instead.

Closes #2163.