Improve conda-forge installation documentation.

[joaander]

Description

Recommend that users request specific versions of hoomd. Also make it more explicit that a properly configured conda-forge environment is required.

Motivation and context

On one systems I recently tested, conda install hoomd offers hoomd 2.9.7. Some users have reported similar behavior on the mailing list.

How has this been tested?

I tested installing both CPU and GPU hoomd builds on a mambaforge environment on NCSA delta. mamba install resolves the environment very quickly, then often sadly fails with an error in the install phase. The conda install commands in the docs work, though they require several minutes of iteration through "Solving environment":

conda install "hoomd=4.0.0=*gpu*"
Collecting package metadata (current_repodata.json): done
Solving environment: unsuccessful initial attempt using frozen solve. Retrying with flexible solve.
Solving environment: unsuccessful attempt using repodata from current_repodata.json, retrying with next repodata source.
Collecting package metadata (repodata.json): done
Solving environment: done

Change log

Fixed:

* The recommended conda install commands find the documented version.

Checklist:

• I have reviewed the Contributor Guidelines.
• I agree with the terms of the HOOMD-blue Contributor Agreement.
• My name is on the list of contributors (sphinx-doc/credits.rst) in the pull request source branch.

[tommy-waltmann]

Why doesn't conda install hoomd install the newest version of hoomd? Is there something that can be done about that upstream?

[joaander]

Why doesn't conda install hoomd install the newest version of hoomd? Is there something that can be done about that upstream?

conda's solver is very complex and often finds solutions that the user does not intend. There are many issues posted for related problems for both vanilla conda:
https://github.com/conda/conda/issues?q=is%3Aissue+outdated+
and mamba: conda/conda-libmamba-solver#64

It is far beyond the scope of my development effort to instruct these solvers to read the user's mind. Perhaps chatGPT is the solution. The best I can do is offer users instructions on how to explicitly ask the solver for the intended specific solution.