Add zero-electron species to gaussian dataset

[msricher]

Can I get your opinions on this, and if it's OK, I will add it for the other datasets as well.
(Currently the new test fails because the .msg file is not in the atomdbdata repo.)

[gabrielasd]

Yes, this seems a good solution to me give that making the change at the level of the atomdb.load function isn't as simple as it seemed. Thanks @msricher

Btw, I've compiled the proton using the updated run script; let me know and I'll upload it here (under test folder) so the tests can be ran.

[marco-2023]

Looks good to me.

[gabrielasd]

Just one comment, when I tried running the promolecule test for LiH, it failed; my impression is that the problem was on the side of the promolecule module, not that the zero-electron species was being generated incorrectly.

I'll upload the $H^+$ species so that thais can be verified. Maybe the decision is still to merge this change, and fix separately the issue with promolecule (if there is one, I may be wrong where the problem is)

[gabrielasd]

@marco-2023 I just added the proton file

Also, I had to comment L150 on the gitignore file to track this file added under test/data/gaussian/db. Maybe be should keep this line as it used to be (atomdb/datasets/*/db/) so we can track changes to the test msg files?

An here is the error I get when running the promolecule test module:
`
# Check that spin number and multiplicity are recovered
if mults is not None:

      assert np.allclose(sum(np.sign(m) * (abs(m) - 1) for m in mults), promol.nspin(), rtol=1e-7)

E assert False
E + where False = <function allclose at 0x7f6c9e262830>(np.int64(1), 0.5, rtol=1e-07)
E + where <function allclose at 0x7f6c9e262830> = np.allclose
E + and np.int64(1) = sum(<generator object test_make_promolecule.. at 0x7f6c798ee740>)
E + and 0.5 = nspin()
E + where nspin = <atomdb.promolecule.Promolecule object at 0x7f6c798e95e0>.nspin

atomdb/test/test_promolecule.py:212: AssertionError

`

[msricher]

I added a special case for computing intensive properties where zero-electron species may be present. It just calls out to _intensive_property_exclude_zero(...) instead of _intensive_property(...).

[msricher]

Now only the nist_data tests fail, which don't concern this PR.

[marco-2023]

@msricher I think the fix is good.
Regarding the failing tests, I don't know why they are failing, all tests pass on my pc after the merge.

[gabrielasd]

I think it was best to only exclude the .msg files under datasets from being tracked; imo we do need the ones under the test module.

Aside from this I also agree with the changes in this PR.

[gabrielasd]

@marco-2023 @msricher
I'll open a separate issue for the NIST tests that fail; to me there seem to be two (possibly related) sources of error

• a change in the value of some constants from scipy.constants module for python versions > 3.9. For the AMU variable we use, the difference between versions is in the order of $10^{-6}$, which is roughly the error in the assertions that fail.
• what is being returned by the attribute atmass of the instance of Species? The value of Elements[element].mass or that from Species._data.atmass which is likely compiled with python 3.9?

[PaulWAyers]

@marco-2023 @msricher and @gabrielasd is this pull request ready to merge now?