Fix/pitch shift

[stephenhensley]

This should end up fixing #162

The negative transposition was almost 2x the ratio that it was supposed to. So that none of the semitones were accurate when set below 0. This has been fixed.

Making the phasor accept negative frequency let's us remove any of the weird "is it going up or down" logic, and just have the phasor flip when the ratio is derived from a negative transposition.

This does remove the pre-generated lookup for the semitones. We can replace it with a more robust lut at some point to get this to be a bit more efficient.

Also moved the one global define into the class as a member constexpr

And added a secondary argument (defaulting to false) to lock transpositions to semitones.

[stephenhensley]

This isn't quite ready. Checked the "fun" parameter, and it is quite fun when tranpsosing up, but when transposing down it just adds a bunch of noise. So we probably need to flip the direction the modulation happens in when transposing below 0

It also does nothing when set to original pitch,

[xconverge]

This would also fix #201

[xconverge]

@stephenhensley is there more work to be done for this? I wanted to open a PR to templatize or parameterize the buffer size but want it based off of this PR. I have run this PR a decent amount and it seems pretty solid

[stephenhensley]

I don't think there's anything immediate left to do here.
At the time, I think there was still some potential clean up, or testing I wanted to do.
But I have used this in a few projects without noticing any problems, and it does resolve #201