Atom energy fitting

[cgrambow]

Motivation or Problem

We should fit atom energies so that levels of theory that produce incorrect atom energies can be used to compute thermochemistry.

Description of Changes

Added AE and AEJob classes that require a dictionary of species labels with single-point electronic energies (computed using the experimental geometries available in the reference database). For example,

species_energies = {
    'Dihydrogen': -1.1607014438,
    'Dinitrogen': -109.5502634582,
    'Dioxygen': -150.3479925723,
    'Disulfur': -796.325655201,
    ...
}

The list of species that has to be provided is available in SPECIES_LABELS in arkane/encorr/ae.py. Atom energies can be fitted by using the ae() function in Arkane input files as described in the updated documentation. Corresponding database PR is ReactionMechanismGenerator/RMG-database#415.

Testing

Added unit tests and checked that it works well for wB97M-V.

Reviewer Tips

Check the logic of the linear regression and maybe run an Arkane job to fit energies.

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (master@36a90a2). Click here to learn what that means.
The diff coverage is 74.80%.

@@            Coverage Diff            @@
##             master    #1950   +/-   ##
=========================================
  Coverage          ?   46.40%           
=========================================
  Files             ?       87           
  Lines             ?    22973           
  Branches          ?     5974           
=========================================
  Hits              ?    10660           
  Misses            ?    11169           
  Partials          ?     1144           

Impacted Files	Coverage Δ
arkane/encorr/ae.py	73.45% <73.45%> (ø)
arkane/encorr/reference.py	69.88% <81.81%> (ø)
arkane/encorr/bac.py	76.42% <100.00%> (ø)
arkane/encorr/corr.py	62.22% <0.00%> (ø)
rmgpy/tools/data.py	71.42% <0.00%> (ø)
rmgpy/ml/estimator_test.py	100.00% <0.00%> (ø)
rmgpy/data/kinetics/common.py	69.03% <0.00%> (ø)
rmgpy/pdep/__init__.py	100.00% <0.00%> (ø)
arkane/kinetics.py	12.24% <0.00%> (ø)
rmgpy/tools/simulate.py	79.24% <0.00%> (ø)
... and 80 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 36a90a2...9c85cc6. Read the comment docs.

[amarkpayne]

Looks good, thanks for implementing this! I looked over the code for fitting again, and all of the math checks out. I'll go ahead begin a PR in ARC to utilize this functionality.

[amarkpayne]

The energies are just floats, right? We should specify that the unit needs to be hartree

[cgrambow]

I mention the units in the notes above. I'll add it here again.