pinning numpy<2

[vineetbansal]

addressing issue 229

[vineetbansal]

@simozacca - this is ready to be reviewed now (should be squashed and merged).

The only major difference here is that we're using ruff instead of blue+flake8 combination as the formatter and linter. Essentially the problem is that if the linter and formatter are separate tools, then they have conflicting opinions on what is "proper" and it becomes a pain to maintain. Also flake8 is becoming somewhat of a pain to work with because its not a single binary (and has its own dependencies to worry about).

The move to ruff allows us to test on python up to 3.12 easily, as we've done here.

Once this is merged, I'll tag a v2.1.0 and update the recipe on bioconda to work against the new release.