Switch CI & Documentation Build to Micromamba from Conda for Reduced Execution Time

[JacksonBurns]

Motivation or Problem

Resolves #2365

Description of Changes

Changes the GitHub CI to use micromamba instead of conda, which resolves the RMG environment in ~2 mins instead of 30+

Testing

Successfully ran on this fork.

Reviewer Tips

The fact that it runs is perhaps sufficient evidence. Might want to do a line-by-line comparison of how the depdencies were resolved to ensure that versions are consistent with conda.

The changes, if good, will also need to be ported into the documentation builder.

[JacksonBurns]

The debug level logging in the call to provision-micromamba should also be removed once the rest of this looks operational.

[codecov]

Codecov Report

Merging #2366 (44bfa20) into main (a8c4127) will decrease coverage by 0.68%.
The diff coverage is n/a.

❗ Current head 44bfa20 differs from pull request most recent head aa97e23. Consider uploading reports for the commit aa97e23 to get more accurate results

@@            Coverage Diff             @@
##             main    #2366      +/-   ##
==========================================
- Coverage   48.25%   47.57%   -0.68%     
==========================================
  Files         110      110              
  Lines       30616    30612       -4     
  Branches     7982     7982              
==========================================
- Hits        14774    14564     -210     
- Misses      14298    14532     +234     
+ Partials     1544     1516      -28     

Impacted Files	Coverage Δ
rmgpy/qm/molecule.py	37.55% <0.00%> (-43.27%)	⬇️
rmgpy/qm/mopac.py	26.43% <0.00%> (-41.38%)	⬇️
rmgpy/qm/qmdata.py	52.77% <0.00%> (-27.78%)	⬇️
rmgpy/qm/main.py	51.61% <0.00%> (-13.71%)	⬇️
rmgpy/rmg/reactors.py	19.95% <0.00%> (-1.04%)	⬇️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[JacksonBurns]

I also propose that we run the CI on a schedule using GitHub actions scheduling. There have been a couple instances in the past couple weeks where the CI was failing on new PRs even when it was passing on main, only because it had not been run in months.

We could do it with this simple addition to the script:

on:
  schedule:
    # * is a special character in YAML so you have to quote this string
    - cron:  '0 0 8 ? * MON,TUE,WED,THU,FRI *'

which would run the CI every Monday-Friday at 8 am UTC (3 am EST)
Please let me know your thoughts @mjohnson541 @xiaoruiDong @kspieks

[JacksonBurns]

I have added the mamba build to the documentation action, as well as the runner schedule to the CI action.

I did a comparison between the environment resolved by micromamba and conda (see it for yourself here and it resolves almost everything to be the same, with the exception of some operating system level packages that I assume are a function of the action being run on a different machine + VM at GitHub every time. All of our 'major' dependencies (those that we maintain or use heavily) are resolved to the correct major versions.

Tests are still passing and this cuts ~30 minutes off CI and Documentation build.

[hwpang]

Why do we reset log level to warning here?

[hwpang]

Can you squash this with the commit that set it to debug?

[JacksonBurns]

Debug writes more to stdout than is useful for any future test running (see this run with debug) -- I only turned it on to deal with the issue of channel-priority. The default log level is warning

[hwpang]

Could you squash this commit with the previous one?

[hwpang]

Thanks for the PR! Overall seems good with a few comments. Several of the commits can be squashed together. I check the fork, and it looks like it fails Trigger RMG-tests?

[JacksonBurns]

Thanks for the PR! Overall seems good with a few comments. Several of the commits can be squashed together. I check the fork, and it looks like it fails Trigger RMG-tests?

I left a quick comment about the logging level. Will squash after it is resolved.

That action fails on forked repositories regardless of content. See here, where I changed nothing in the source code and it still fails at the same location.

[JacksonBurns]

Moved to #2386