Allow destructuring assignment on fixed-length vectors

[tredoe]

This expression looks refutable to me, but the compiler shows an error:

let [a, b] = [1, 2];

[bstrie]

Historically the problem here is that an expression like let [a, b] = ... isn't irrefutable in the general case. Even something like this:

let [head, ..tail] = foo();

is refutable in the case where the result of foo() is an empty vector.

However, I wonder if this can be made to work only in the case of functions that return fixed-length vectors. If we know that the return type of foo() is [T, ..N] for any N > 0, this should be irrefutable, correct? Would that be useful?

[bstrie]

Updated issue with new title.

[graydon]

reproduced 2013-06-11, tagged. nominating (weakly) for feature complete. not a backwards compat issue.

[bstrie]

Rolling this into #7784 and closing.