Fix character encoding conversion issues

[grorp]

Continuation of #221. Fixes #216 and fixes luanti-org/luanti#13646.

The Irrlicht string conversion functions are not broken, they are just the wrong kind of conversion functions: Irrlicht uses wide <-> multibyte conversion functions in places where it should actually use wide <-> UTF-8 conversion functions. This usually works because the locale-defined multibyte encoding is UTF-8 in many environments, but sometimes, it doesn't.

This PR introduces new wide <-> UTF-8 conversion functions, utf8ToWString and wStringToUTF8, and uses them where appropriate. wStringToMultibyte is not used anymore and has been removed, multibyteToWString has been kept because it is still used in one place.

EDIT: btw, "AutomatedTest" doesn't even compile on Windows.

To do

This PR is Ready for Review. Before merging, 69622fe should be reverted.

How to test

Compile Minetest on Windows with and without SDL. Set a weird combination of language settings. Verify that Unicode text is displayed/copied/whatever-ed correctly.

[numberZero]

Can a stringc contain embedded nulls? If so, this shouldn’t delegate but use another from_bytes:
conv.from_bytes(source.begin(), sorce.end())

[grorp]

This seems to suggest it can't:

irrlicht/include/irrString.h

Lines 789 to 798 in 5b2f192

	//! verify the existing string.
	string<T>& validate()
	{
	// truncate to existing null
	u32 len = calclen(c_str());
	if (len != size())
	str.resize(len);
	return *this;
	}

[sfan5]

rebase mistake?

[grorp]

I temporarily disabled the new driver to make CI pass: 69622fe
I'll revert that commit and rebase this PR now that #250 is merged.

[sfan5]

Merged.

EDIT by grorp: as commits 93eebed and c766c3a

[grorp]

Thanks!