Ensure save_dataframe writes atomically by default

[jbweston]

This PR makes it so that 'save_dataframe' writes atomically by default.
This is important for maintaining data integrity during an adaptive scheduler run.

This adds a dependency on the 'atomicwrites' package.

Closes #181.

[codecov]

Codecov Report

Attention: 9 lines in your changes are missing coverage. Please review.

Comparison is base (cd3e66c) 72.88% compared to head (97f86a1) 73.46%.

❗ Current head 97f86a1 differs from pull request most recent head db3fdec. Consider uploading reports for the commit db3fdec to get more accurate results

Files	Patch %	Lines
adaptive_scheduler/utils.py	74.28%	9 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #182      +/-   ##
==========================================
+ Coverage   72.88%   73.46%   +0.58%     
==========================================
  Files          22       22              
  Lines        2482     2480       -2     
==========================================
+ Hits         1809     1822      +13     
+ Misses        673      658      -15     

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

[basnijholt]

There was some issue with atomicwrites being yanked from pypi by its maintainer.

[basnijholt]

Can also be done like this: python-adaptive/adaptive#353

[jbweston]

Can also be done like this: python-adaptive/adaptive#353

Does that handle calling sync and everything?

[basnijholt]

Not sure about all the details. I just remembered the panic a few years back when the package was yanked, hehe.

And python-adaptive/adaptive#352

but adding it here again is fine

[basnijholt]

Python 3.9 CI job seems to consistently get stuck?

[jbweston]

Python 3.9 CI job seems to consistently get stuck?

Yes, no idea what is causing this, and cannot reproduce locally.