Refactor attribute_of_binary to reduce instantiation cost

[yaito3014]

fixes #56

[cppwarningnotifier]

Environment				C++23	C++26
x4	Clang	21	Debug	✅success	✅success
			Release	✅success	✅success
	GCC	14	Debug	✅success	✅success
			Release	✅success	✅success
	MSVC	2022	Debug	✅success	✅success
			Release	✅success	✅success
		2026	Debug	✅success	✅success
			Release	✅success	✅success

[saki7]

I'm not sure why, but it seems it's getting worse in some cases.

---

---

---

---

Maybe these lines should also be lightened:

x4/include/iris/x4/operator/alternative.hpp

Lines 43 to 47 in aa3a7f1

	using attribute_type = std::conditional_t<
	is_both_same_attribute,
	std::type_identity<typename parser_traits<Left>::attribute_type>,
	traits::attribute_of_binary<iris::rvariant, Left, Right>
	>::type;

This was part of the recent PR #55, and the PR added another complexity in the attribute type, I think.

P.S. The refactored code itself is very good, more readable for humans so it's better if we could keep the simple structure and solve the IDE heaviness at the same time.

[saki7]

Still not working, but the vector<bool> case improved, so it definitely has something to do with template complexity, I guess.

To further reduce the instantiation depth, I think we should seriously start considering to drop the _t type aliases from implementation-detail metafunctions. I often see the standard library implementations do not define the _t variants for internal metafunctions, they just use ::type. From my experience it has some impact on the type expansion because type aliases appears to be treated like distinct template classes.

---

---

---

[saki7]

Fixed.

---

---

---