This release includes some major (breaking) changes to the `amplitude` API, but the benefit is a speed increase of around 60% according to some benchmarks.
First, we got rid of a few unnecessary allocations which were happening quite a lot inside the main loop. Second, we got rid of a match statement which was also happening multiple times inside that same loop. This theoretically improves branch prediction and maybe even some cache locality, which speeds things up a ton!
Note that the main changes for developers (I figure I'm the only one at this point, but this is for posterity) are that `AmpOp` was removed, we can just return `Amplitude`s and interact with those. The rest of the breaking changes are a bit behind the curtain, but there are also par_ and regular versions of most of the end-stage computation methods just to allow for profiling without `rayon`. This shouldn't change the Python API at all.