[extension types] Adjust the paragraph about Object members: They do not need special treatment

[eernstg]

The extension type feature specification used to have a paragraph about invocations of members of Object (hashCode, operator ==, etc). It stated that the signature was taken from Object?. This is not consistent with a situation like the following, because all other non-extension type members would be treated according to the signature known from the traversal of the superinterface graph.

This PR changes the given paragraph to commentary, and changes the text to simply say that members of Object receive exactly the same treatment as all other non-extension type members. This yields the following behavior:

class A {
  void foo() {}
}

class B implements A {
  @override
  void foo([int i = 0]) {}

  @override
  String toString([bool whatever = true]) => super.toString();
}


extension type E1(B b) implements A {}

extension type E2(B b) implements E1, B {}

void main() {
  var e2 = E2(B());
  var e1 = e2; // OK, `E2 <: E1`.
  
  e2.foo(42); // OK, signature is `void foo([int])`.
  e2.toString(true); // OK, signature is `String toString([bool])`.
  
  e1.foo(42); // Compile-time error, signature `void foo()`.
  e1.toString(true); // Compile-time error, signature `String toString()`.
}

Note that we need to take the combined member signature of foo in E2: E1 contributes the signature void foo() and B contributes the signature void foo([int]), and the combined member signature is then void foo([int]). The signature for toString in E2 is computed in the same manner (as mentioned in the title: there is no need to special-case members of Object).

[lrhn]

Ship it!
(We need tests. Lots of test.)

[eernstg]

@sgrekhov, could you take a look at the co19 tests in this area, possibly adjusting existing ones and adding new ones?

[sgrekhov]

dart-lang/co19#2304