Fix enable_ifs for map formatter: map formatter requires std::pair formatter but never uses it

[BrukerJWD]

I am referring to L455 in the formatter definition if is_map<T> is true:

fmt/include/fmt/ranges.h

Lines 449 to 456 in 5682338

	template <typename T, typename Char>
	struct formatter<
	T, Char,
	enable_if_t<detail::is_map<T>::value
	// Workaround a bug in MSVC 2019 and earlier.
	#if !FMT_MSC_VERSION
	&& (is_formattable<detail::uncvref_type<T>, Char>::value ||
	detail::has_fallback_formatter<detail::uncvref_type<T>,

Please note: T is the map type. We are using detail::uncvref_type<T> here, which gives the type of a dereferenced iterator of T, i.e. a std::pair<const K,V> in case of a std::map<K, V>.
This formatter is never used, though, as key and value are formatted separately:

fmt/include/fmt/ranges.h

Lines 476 to 479 in 5682338

	out = detail::write_range_entry<Char>(out, item.first);
	*out++ = ':';
	*out++ = ' ';
	out = detail::write_range_entry<Char>(out, item.second);

Thus, a formatter for the std::pair is not needed, or more generally speeking: L455 isn't checking the correct thing.

I stumbled upon this with a custom map which isn't using std::pair internally, and doesn't have a formatter for the internal pair type. But this shouldn't hinder this formatter from being used.
Additionally, the pair<K, V> formatter is no guarantee that we have formatters for K and V of the map (although it probably would be bad code not to use them in the pair formatter).

[vitaut]

Thank you!