Update analyse_freenrg to use pymbar 4

[fjclark]

Is this pull request to fix a bug, or to introduce new functionality?

Since it was a quick fix, I've updated analyse_freenrg to use the pymbar 4 API. I tested on two sets of free energy data before and after the changes and got identical results.

• I confirm that I have merged the latest version of devel into this branch before issuing this pull request (e.g. by running git pull origin devel): y
• I confirm that I have added a changelog entry to the changelog (we will add a link to this PR as part of the review): y
• I confirm that I have permission to release this code under the GPL3 license: y

Suggested reviewers:

@chryswoods, @lohedges

[chryswoods]

I approve of all of these changes. Thanks for updating the requirements_bss and try_import lines to the latest versions :-)

[chryswoods]

@jmichel80 - is this something you are happy to merge? Do you want to run more tests? This is a change in functionality, so would go into 2023.4.0.dev for a development period, and would only be properly released and supported once 2023.4.0 is released at the end of September. I am aware though that this is close to the workshops...

[jmichel80]

Could we support both pymbar versions easily for the time being ? The default would be pymbar 4 but it would make it possible to revert to using BSS in an environment with pymbar3 in case we run into issues.

[fjclark]

Could we support both pymbar versions easily for the time being ? The default would be pymbar 4 but it would make it possible to revert to using BSS in an environment with pymbar3 in case we run into issues.

@jmichel80 I've updated the code to be compatible with either version of pymbar. I've tested with pymbar greater than, and less than 4 on the same data (bound vanish leg of a MIF ABFE calculation) and got identical results.

[chryswoods]

Thanks @fjclark - I've just merged in feature_align and it is going through GitHub Actions. Once that is done I will update this PR to match devel, and will then merge it in. Thanks for doing the work to port to both versions of pymbar simultaneously. This is super helpful and will make it much easier to build flexible conda packages of sire. I think this will enable us to drop pymbar as a dependency and use try_import to bring it in at runtime. The right version will be brought in based on the environment, or the user can pre-install the version they want (e.g. I suspect we will use the older version on the notebook server for the workshop, but will look to use a newer version later in the year).

[fjclark]

Great, thanks very much @chryswoods. No problem at all.

[chryswoods]

Thanks - this will take a few hours to build. It should be in the devel binaries by tomorrow.