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

Issue #444

This is not obnoxious now that logging is off by default

Eventually extensions will probably need access to the parser again, but it'll
be in a different form.

After this we can remove the parser from the syntax extensions, at least for
now.

Issue #444

Issue #444

Introduce an ext_ctxt record to provide a span_err method for use while
expanding syntax extensions. Hopefully it will be useful for other things.

This is accepted everywhere, since just passing a pointer is safe.

So that the tinderboxes can go green again. Patrick, please look into
this at some point.

It passes thanks to the previous two commits.

Generate an LLVMUndef value of the proper type for an argument of
type _|_. Otherwise we get an LLVM assertion failure.

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