Explicit float and integer formatting

Currently, formatting all floating-point values uses the shortest possible representation by default (as implemented in /src/libcore/fmt/float.rs. This is implemented such that floating-point values with no fractional parts are displayed as integers, e.g. 13.0f64 becomes formatted as "13". This also means that the only way to currently always achieve a formatting that always displays the decimal point is to give a specific precision.

So as not to introduce any incompatibilities, I'd suggest introducing two new format types and traits associated with them as such:

• d => Decimal
• f => Fractional

The Decimal trait should rounded to the nearest integer of the value and represent it as an integer without any decimal point.
The Fractional trait should always display the value with a decimal point. If no precision is given, it produces the fewest digits after the decimal point needed to display the exact value. If precision is given, it behaves the same as it currently does without any specified format type.

This can be technically achieved with relative ease for the f32 and f64 types as the float_to_decimal_common function in /src/libcore/fmt/float.rs can be changed to accept a minimum decimal point parameter which can be passed to the call to to_shortest_str to be used for the Fractional trait.

For the Decimal trait, the float_to_decimal_common function can be modified further to accept an argument to display as integer with no fractional part.

Read the proposal and the referenced src/libcore/fmt/float.rs, starting with float_to_decimal_common and to_shortest_str. Trace how existing floating-point formatting handles precision and format types. Done means the proposed Decimal and Fractional behaviors are defined without breaking existing formatting, with the f32 and f64 paths covered.