Add support for destructuring vectors in match expressions

[ghost]

Resubmitting #4091 as I cannot seem to be able to change the target branch in the original PR. I squashed the commits as there isn't much of a logical separation possible and some of them were in half-broken state.

It'd be great if we could get this in for 0.5 if it's not too late. I'm stil considering @erickt's [a, ..b, c] proposal and wonder what other people think? As we're rejecting that syntax now, we can still be forward compatible with just the [a, ..tail] variant, though, so this could be added later on.

[brson]

This is a very nice looking patch.

[brson]

Since this was written, the default binding mode changed from the old implicit by-ref to copy, so I hope all these test cases are still valid (if not they can be changed to ref. It would be nice to have test cases using all the binding types: ref, copy and move.

[ghost]

@brson The tests are still passing except for ./compile-fail/alt-vec-illegal-tail-loan.rs due to what seems to be #3243. It should probably be xfailed for now unless it's actually an oversight in this patch...

[graydon]

I'll merge this presently, just wanted to check .. this is substantial enough that we should probably have a contributor agreement. Have you filled one of those out?

[ghost]

@graydon No, I have not... Is this the one you have in mind: http://www.mozilla.org/hacking/committer/? I'll take a look.

[ghost]

@graydon I submitted the agreement via e-mail today.

[ghost]

I rebased and made some more changes to make the exhaustiveness/reachability checks cleaner.

The lifetime problems with destructured tails are clearly an oversight on my part, I hadn't looked into the memory categorization pass at all.

[graydon]

Excellent. Contributor forms received, landed this in 6c83fe4 , thanks so much!