Fixed issue with corr_terms and confidence envelopes, and more

[tomasstolker]

Hi @wbalmer and @gotten,

This PR contains a bit more changes then I initially planned to include so please have a look if you agree with these changes 😉.

There are two issue that I fixed:

• When corr_terms include both True and False, it was only including the uncorrelated terms in likelihood function because of if ~corr_terms.all():
• The 1 sigma envelopes were not correctly implemented from what I understood. This I have fixed by taking 200 samples from the posterior and calculating the percentiles.
• Added the samples (instead of envelopes) to the deltaRA-deltaDEC plot

A couple of changes to the parameters:

• The step_size can be set with generate_plots to speed up trackplot
• Removed the jd_tt parameter (see issue #19) and instead added ref_epoch to generate_plots. When set to None (=default), it uses the mean date form the input astrometry. Change the days forward and backward to 3 years.
• Added fileprefix parameter to system class (e.g. for the dynesty files). Perhaps we could use a single fileprefix for all methods?

And some more updates:

• Simplified the extraction of samples by using samples_equal from dynesty
• Added Github Actions workflow
• Updated .gitignore and init
• Added unit tests to the tests folder

Great that the package is now available on PyPI! I noticed that the name has been changed to backtracks. I have changed the name also in the folder and filenames. @wbalmer probably best to also change the name of the Github repo?

About the package structure: I would suggest to simplify it to just having a backtracks folder (like before) instead of embedded in src. I think that is a quite common package structure, in particular because there are only a few files. But let me know in case you disagree then I will change it back!

Also, I would recommend to not add any binary files, so only include text files. Any changes to a binary file will store a full copy of the file so quickly increasing the repo size. I just checked and it is 70 MB already when cloning from Github. I have therefore removed all binary files that seemed not needed to include. The unit tests run in a few minutes so I think it is fine to run the actual fit instead of including a precalculated fit. If you want to run them, just use pytest in the main folder.

Finally, I tested the installation with Python 3.11 (with pip install .) and that seems to work fine as well!

[tomasstolker]

Hereby an example of the trackplot for HD 131399:

[wbalmer]

This is really excellent, thank you so much for clarifying the package structure, the pytests, etc. I'll review in just a few minutes and merge.

[tomasstolker]

Sorry for the many commits... It took me a while to figure out how to setup the Actions workflow because installing orbitize caused an error. You can squash the commits if you decide to merge. But let me know if you have any comments!