Implement is_structured_binding metafunction

[katzdm]

The is_structured_binding metafunction was added in P2996R3.

[BaLiKfromUA]

Planning to implement this function.

Let me know if there are any objections.

[katzdm]

@BaLiKfromUA Sounds great.

A careful review through the content of dcl.struct.bind could be worthwhile - In particular, note that a structured binding declaration proceeds by first introducing a "variable with a unique name e", followed by the declaration of n-many "names" which may or may not be variables; it is these "names" that are the structured bindings, rather than the implicit "variable with a unique name" (which we should avoid producing a reflection of at all).

Note also that there are three cases:

1. The initializer for the structured binding declaration has array type.
2. The initializer for the structured binding declaration is tuple-like (as described here).
3. The initializer for the structured binding declaration is a struct (meeting certain constraints).

And that the structured bindings introduced for cases (1) and (3) are said only to be "lvalue(s) that refer to" subobjects of the implicit "variable with a unique name" - in particular, they are not variables (but the names introduced by case (2) are variables). The is_variable metafunction should behave accordingly.

This godbolt, and the AST structure surfaced therein, demonstrates all three cases; it may be helpful with both the implementation and for writing tests. Please check that both is_structured_binding and is_variable return as expected for all three cases. It may or may not also be necessary to implement special cases for some other metafunctions (e.g., type_of? also maybe extract?); feel free to do so in the same PR, or in a follow-up PR, according to your preference. Please also add tests to cover any such "interesting" special cases, as you see fit.