Updated version specification for nlopt to try to resolve issues

[jmwright]

A new version of nlopt (2.9.0) was released on on PyPI on November 16th. I have updated setup.py for this version and newer to try to resolve the issues that some users are having.

https://pypi.org/project/nlopt/

I did not try to pin the version for conda because it does not seem to be needed. Let me know if someone thinks I need to do that as well.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 95.33%. Comparing base (2629bf9) to head (c19acf5).
Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1718   +/-   ##
=======================================
  Coverage   95.33%   95.33%           
=======================================
  Files          27       27           
  Lines        6774     6774           
  Branches     1010     1010           
=======================================
  Hits         6458     6458           
  Misses        193      193           
  Partials      123      123           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[adam-urbanczyk]

Thanks, LGTM!

[lorenzncode]

Thanks @jmwright !

[kmarchais]

Hello @jmwright, the nlopt 2.9.0 package is not available for macos x86_64.
It switched from x86_64 to arm between 2.7.1 and 2.8.0.

What kind of issues do this PR solve?
Do you think it is possible to loosen this constraint at least for this specific platform and architecture?

[jmwright]

@kmarchais 2.7.1 was not working with numpy >= 2. If I relax the version constraint for MacOS x86_64 things will start breaking with CadQuery master, the upcoming 2.5 release, and beyond. I wonder why nlopt dropped Intel Mac support. There are still a lot in circulation, and it's still a standard runner in most CI systems. Is using conda an option for you?

[kmarchais]

Thank you for the info, I understand. We will use it with this constraint then. Using conda is an option, yes. Ideally, I would have preferred to keep our library compatible on Intel Mac with pip.