Clean up requirements

[mslw]

Currently we have requirements-devel.txt and pyproject.toml. The latter has tool.hatch.metadata.dependencies and several tools.hatch.envs.*.extra-dependencies.

The pyproject.toml has a comment which recognizes the problem, and suggests that we would first need a released version of datalad-core: This section is required to allow the direct refernces to the datalad_core in the dependencies section below and in tool.hatch.envs.hatch-test.extra_dependencies. Remove the section tool.hatch.metadata when datalad_core is available on PyPI. -- but I thought it would be worth surfacing that in the issue tracker.

As it stands right now, pip install would ignore the dependencies in pyproject.toml (as implied by the comment, a top-level dependencies section would be needed for that). The readme currently recommends pip install -r requirements-devel.txt, but that has more than strictly needed just to try it out (coverage, hatch, hypothesis, sphinx...). FWIW, I have so far used pip install annexremote datasalad datalad-core datalad-next toml followed by pip install -e ..

Also - shouldn't the DataLad package names in pyproject.toml have hyphens instead of underscores? That's how they appear on PyPi, although the package names use underscores.

I think it would be ideal if we ended up the dependency specification in pyproject.toml only.

I am not yet familiar enough with how hatch dependencies / envs / extra dependencies interact with pip. Do we have to provide dependencies for hatch-install and pip-install separately? Do we plan to have two installation methods (pip & hatch) in the README?