Remove boost dependency

[LordAro]

I'm not going to claim it works, needs someone with far more knowledge of grfcodec to me to actually test it. But it does compile, so...

• Replaces boost::for_each with actual ranged for loops and algorithms
• Replaces boost::bimaps usage for NFO escapes with std::vector<std:pair>
  • I'd be amazed if there are enough escapes to give the bimap structure any benefit
• Replaces boost::gregorian with std::chrono
• Replaces some godawful preprocessor-generated classes with some inheritance (I'm least sure of this)

[rubidium42]

I support the removal of boost.

[rubidium42]

Why not convert this into a [[noreturn]] function while you're at it? Something akin:

[[noreturn]] static inline void INTERNAL_ERROR(std::string_view var, const auto &val, const std::source_location location = std::source_location::current()) 
{
	IssueMessage(0, INTERNAL_ERROR_TEXT, location.file_name(), location.line(), _spritenum, var, val, location.function_name());
	assert(false);
	exit(EFATAL); 
}

[LordAro]

I did try that initially, but ran into the issue that var is not always a string and I wasn't willing to change all the callsites. That said doing so is probably smaller than all the other changes I ended up making

[rubidium42]

This triggers my spidey-sense:

if (foo) VERIFY(bar, baz);
else std::println("foo: {}", foo);

Is going to print foo: true. I know this situation doesn't exist in the code, but it might become surprising at some point. The previous code did not open that hole.

[rubidium42]

For what it's worth, I reckon extra is equivalent to either 0 or local_days(SINCE_1920). It would be more logical if extra became something like epoch, with 0 in the _D_ branch and SINCE_1920 in the _W_ branch. And then just subtracting epoch in this return.