Equality operator allowed for external function types

[nishant-sachdeva]

Closes #12411

[chriseth]

Looks good so far! Please also experiment a bit with comparing function pointers of non-identical type, i.e. compare view, pure, payable and non-payable function pointers to different function pointers.

[ekpyron]

As far as I can see, this mainly needs better tests now.

[ekpyron]

The antlr grammar test failure is probably actually a bug in the antlr grammar - maybe I'll take a look at that myself tomorrow to fix it. The rest of the test failures as far as I can see are due to a mismatching source location for one of the errors, so running isoltest locally and updating the tests should fix that.

[ekpyron]

#12545 is merged, so this can be rebased.

[ekpyron]

Looks good, but we should still add, resp. add back, some assertions about the stack size of the types in both old and new code generation. (i.e. #12470 (comment) and #12470 (comment))

[ekpyron]

Looking at this once more, it may also have been nice to build this as

Suggested change

	expr = m_utils.externalFunctionPointersEqualFunction() +
	"(" +
	IRVariable{_binOp.leftExpression()}.part("address").name() + "," +
	IRVariable{_binOp.leftExpression()}.part("functionSelector").name() + "," +
	IRVariable{_binOp.rightExpression()}.part("address").name() + "," +
	IRVariable{_binOp.rightExpression()}.part("functionSelector").name() +
	")";
	expr = m_utils.externalFunctionPointersEqualFunction(functionType) +
	"(" +
	IRVariable{_binOp.leftExpression()}.commaSeparatedList() +
	"," +
	IRVariable{_binOp.rightExpression()}.commaSeparatedList() +
	")";

and then have externalFunctionPointersEqualFunction assert that the function type has only two stack slots...

But I'd say we can keep things like they are for now and change this, if we ever want to compare multi-slot function types (with additional gas slots, etc. - which I'm not sure we will, since it's probably not particularly useful).

So instead a simple assertion for the type having sizeOnStack() == 2 will also be fine.

[ekpyron]

In the Yul codegen part, we could probably have spent some more time unifying this with the other cases (e.g. args is unused in the function type path now - we might have defined auxiliary variables containing cleaned up values in all cases and then defined a tuple comparison function as util function instead of the specialized external function pointer comparison function)...
But at a quick glance I see that that would require some more refactoring of the IRVariable mechanism, so I'd say we stay with the current solution for now :-).