Underlying unicode char types, updates to JSON reader

[trueqbit]

Originally i set out to correct a warning about snprintf() requiring an unsigned int when passed a char32_t.

So this PR fixes this warning and more:

• It is better, or even the only correct approach, to use the documented underlying types of C++ Unicode character types for byte-based, arithmetic, or bitwise operations. These are defined to have at least the required bit size, as opposed to a fixed bit size. Also there are platforms where integral types of fixed bit size are unavailable. Specifically: char -> unsigned char, char8_t -> unsigned char, char16_t -> std::uint_least16_t, char32_t -> std::uint_least32_t.
• Fixed a crash in json::basic_reader<>::match() due to accessing the character array past-the-end.
• Fixed non-compiling json::basic_reader<>::match_any().
• Replaced character-pointer logic in json::basic_reader<> with string_view functionality or iterator logic.
• Added noexcept specifier to lambda functions passed to json::basic_reader<>::consume_while().
• Replaced sizeof with std::size() when passing character buffer size.

Additionally:

• The CMake project was built with Linux in mind. I updated it so it works on Windows with Visual C++.
• In addition, I added the compiler option to treat the warning about incorrect order of initialization of members as an error, which I found very useful in other codebases.

[eteran]

Fantastic! I'll take a look at this in a little bit and try to get it merged quickly.

[eteran]

PR looks pretty fantastic, once you address the (very minor) stylistic nit picks, I'm happy to merge it in!