Add concurrency support to autobpm processing

[semohr]

Summary

Currently, autobpm processes tracks sequentially, which can be slow for large libraries. See implementation here.

Motivation

The current sequential approach does not fully utilize modern multi-core CPUs or high-throughput storage. By adding concurrency, autobpm could process multiple tracks simultaneously, reducing overall runtime for large libraries.

References

• Original discussion: Improving autobpm concurrency #6435

Additional Notes

This seems like a straightforward optimization and could significantly improve processing time for users with large libraries.

[jcjordyn130]

Thank you for making this an issue, looks good

[snejus]

Good idea. I personally don't use autobpm primarily because of lack of concurrency and how long it takes to import librosa. I wonder whether we can avoid re-importing librosa in each concurrent process?