no_std support

[CBJamo]

This is a draft to add no_std support. I simply added the #[no_std] directive and fixed compiler errors, swapped all floats to f32, and added micromath to provide float operations. The tests still pass, and the example encode/decode program works the same. I have not tested against the original codec2 implementation, though I would be quite surprised if it didn't work.

[scriptjunkie]

This seems good, move out of draft when you're ready.

[CBJamo]

I did a little additional testing, using the C reference, this no_std, and the existing implementation to check if there are differences in the outputs. Here are all the versions in an audacity project.

I don't know audacity well enough to do anything but compare them visually, but to my eye the the no_std encode must produce similar output to the reference and existing impls, but the no_std decode introduces more noise. I can't tell the difference by ear, however.

In my mind that's an acceptable outcome, but if you want exactly the same performance I'll dig deeper. I suspect the micromath approximations only effect decoding and that switching to a slower but more accurate approach would fix the perf difference.

[CBJamo]

I feel like more testing should be done, but moving micromath/no_std behind a feature feels like an ok middle road. FYI, I'm going to look into getting rid of alloc as well, though I think that will be difficult.

[Ralith]

@scriptjunkie friendly ping? This would be nice to have!

[CBJamo]

Thanks for the review and good catches Ralith, all fixed.

[scriptjunkie]

Pardon the long delay!