Cannot override float processing

[lexxmark]

Tried to override format for standard types according to documentation:

`
using arg_formatter = fmt::arg_formatter<fmt::back_insert_rangefmt::internal::buffer>;

class na_arg_formatter : public arg_formatter {
public:
na_arg_formatter(fmt::format_context &ctx, fmt::format_specs *spec = nullptr)
: arg_formatter(ctx, spec),
m_na_formatter(ctx)
{}

using arg_formatter::operator();

auto operator()(double value) {
	if (IsNA(value))
		return m_na_formatter("NA");
	return arg_formatter::operator()(value);
}

auto operator()(float value) {
	if (IsNA(value))
		return m_na_formatter("NA");
	return arg_formatter::operator()(value);
}

auto operator()(unsigned value) {
	if (IsNA(value))
		return m_na_formatter("NA");
	return arg_formatter::operator()(value);
}

auto operator()(int value) {
	if (IsNA(value))
		return m_na_formatter("NA");
	return arg_formatter::operator()(value);
}

private:
arg_formatter m_na_formatter;
};

inline std::string na_vformat(fmt::string_view format_str, fmt::format_args args) {
fmt::memory_buffer buffer;
// Pass custom argument formatter as a template arg to vformat_to.
fmt::vformat_to<na_arg_formatter>(buffer, format_str, args);
return fmt::to_string(buffer);
}

template <typename ...Args>
inline std::string na_format(
fmt::string_view format_str, const Args &... args) {
return na_vformat(format_str, fmt::make_format_args(args...));
}
`

Float overloading is not working.
Double case is chosen.
na_format("{}", 4.f);

[vitaut]

This is effectively a subset of the issue reported in #1336 - we should stop promoting float to double.

[vitaut]

Fixed in #1360.