Provide no-float implementation of nanoprintf

[ZERICO2005]

compiling nanoprintf without floating point support saves around 2kb. We should probably provide multiple versions of nanoprintf to serve different user needs and size requirements.

I propose adding these 3 implementations. Users can switch between them by defining macros in the makefile

- nanoprintf no-float
- nanoprintf float (default)
- nanoprintf long double

If users want more control, then they can compile nanoprintf.c themselves (which also comes with the benefits of LTO).