More fine-grained feature gates for formatting methods

[mzabaluev]

It would be good to get a subset of the formatting functionality available on no-std targets. On the surface, the format methods can be changed to return alloc::string::String without breaking the current API. Also, #375 discusses formatting into a &mut impl core::fmt::Write, which would require neither std nor alloc.

Consider the feasibility of introducing more fine-tuned formatting feature gates to decouple availability of these methods from the requirement of std:

[features]
formatting = ["formatting-alloc", "std", "itoa/std"]  # full API including format_into
formatting-alloc = ["formatting-core", "alloc"]  # for format
formatting-core = ["itoa"]  # for formatting into &mut fmt::Write

[mzabaluev]

In a future backward-incompatible release, the formatting feature could be redesigned to be orthogonal to alloc/std and enable methods in combination with those:

• #[cfg(all(feature = "formatting", feature = "alloc"))] would enable format;
• #[cfg(all(feature = "formatting", feature = "std"))] would enable format_into;
• #[cfg(feature = "formatting")] would only enable format_into_fmt.

[jhpratt]

The formatting feature currently relies on std, so that change by itself is not breaking. I haven't had time to look at this in further detail.

cc @CryZe as I know you looked into this a bit.

[jhpratt]

As this was predicated on support for fmt::Write, closing for the time being. It may be reopened if necessary once support for fmt::Write lands (there's no timeframe for this).