Add proposal for type inference from method group

[jcouv]

Relates to championed issue: #9007

[jnm2]

Suggested change

	- infering bounds for the `Ui` type parameters in `C<U1 ... Uk>` when given a `C<V1...Vk>` and `C` is a class, struct, interface or delegate type.
	- inferring bounds for the `Ui` type parameters in `C<U1 ... Uk>` when given a `C<V1...Vk>` and `C` is a class, struct, interface or delegate type.

[jnm2]

Suggested change

	It allows the natural type of a method group to contribute to method type inference
	It allows a method group with a unique signature to contribute to method type inference.

Because as written I think the spec would still work even if method groups had no natural type.

[333fred]

Perhaps expand on potential breaking changes that will come from this? IIRC that was a major reason we didn't do this in the first place.

[jnm2]

E.g. the exact inference causes:

static bool P(object o) => true;
static void M<T>(Func<T, bool> f, T value) => Console.WriteLine(typeof(T));

M(P, "hello"); // Today, prints System.String; after proposal, prints System.Object
M(bool (object o) => true, "hello"); // Prints System.Object

[jnm2]

Similarly, this would stop compiling:

static bool P(object o) => true;
static void M<T>(Func<T, bool> f, Func<T, bool> g) { }

M((string s) => true, P); // Compiles now, would behave like the next line after the proposal
M((string s) => true, bool (object o) => true); // Fails to compile

[333fred]

Those are good to include, but neither would have worked before the original lambda changes. The breaks I'm referring to are when an inner scope method suddenly becomes applicable where it wasn't before, such as an instance generic now working instead of going to an extension method, or a derived overload now working where previously it went to a base type.