Allow disabling floating point support

[albaguirre]

Add FMT_USE_FLOAT, FMT_USE_DOUBLE and FMT_USE_LONG_DOUBLE to allow a
user of the library to configure the float types they want to allow.
This is specially useful in embedded environements where code size is
important.

I agree that my contributions are licensed under the {fmt} license, and agree to future changes to the licensing.

[vitaut]

Thanks for the PR. Have you verified that this indeed results in smaller binary code? Could you provide any specific numbers?

[albaguirre]

Thanks for the PR. Have you verified that this indeed results in smaller binary code? Could you provide any specific numbers?

Yeah on our current image, stripped and compiled with -0s using the GCC 4.8 bare metal toolchain (available at https://launchpad.net/gcc-arm-embedded/+download) we save the following:

FMT_USE_FLOAT=1 FMT_USE_DOUBLE=1 FMT_USE_LONG_DOUBLE=0
Saves 3264 bytes

FMT_USE_FLOAT=1 FMT_USE_DOUBLE=0 FMT_USE_LONG_DOUBLE=0
Saves 3696 bytes

FMT_USE_FLOAT=0 FMT_USE_DOUBLE=0 FMT_USE_LONG_DOUBLE=0
saves 12664 bytes

[vitaut]

Do you need partial float support? If not I'd suggest having just one control to enable/disable all floating-point formatting instead of doing it type-by-type.

[albaguirre]

Do you need partial float support? If not I'd suggest having just one control to enable/disable all floating-point formatting instead of doing it type-by-type.

Yes, on some platforms we need single precision float formatting, but not double or long double.

[vitaut]

Looks good in general but let's make it a bit less intrusive.