Some pull requests are failing the balmer tests

[kslong]

When one attempts to merge into dev some pull requests are failing the Balmer test. We need to understand why this is, and we need to have a procedure that checks any tests that will be performed on github earlier in the process.

Note that it is not immediately clear why the Balmer test is failing though I have confirmed that the Polynesian branch does frail the test locally, and that the error is significant

The output from py87h, which passes says:

Array of line ratio relative errors for Balmer series:
 [0.0251518  0.         0.04412924 0.06414727 0.11292285 0.1057267
 0.11958364 0.13849552]

Test passed?: True.

The output from py87i, which fails says:

Array of line ratio relative errors for Balmer series:
 [0.02159244 0.         0.08995894 0.18186688 0.31303101 0.36087883
 0.41898865 0.48026316]

Test passed?: False

Note that the test compares calculated line ratios o the numbers in Osterbrock, so a value of 0.48 means, one deviates by 50% (in one direction or another from Osterbrock.)

The tests is currently set to fail if any of the values printed above exceed a TOLERANCE of 0.3

[kslong]

OK, I understand why py87i is failing the Balmer tests as it currently exists. The balmer_test.pf file uses ML93 ionization prescription, where I had made a change. Previously, in dev, the ML93 prescription solved for T_e in the normal way, which I think was a mistake that had crept in. In ML93, see the end of section 2.2, they set T_e = 0.9 T_r. I discovered we were not doing this for the Tardis comparisons, and this is one reason we were getting poorer Tardis comparisons than previously. So in py87i, I had changed this.

If one changes the ionization prescription to matrix_bb. Then the results of the Balmer test for 87i are as follows:

Array of line ratio relative errors for Balmer series:
 [0.01894222 0.         0.04730428 0.06722004 0.11131465 0.11605118
 0.12662894 0.14755854]

Test passed?: True

which is very similar to what was being obtained with py87h (dev)

Do you @jhmatthews see any reason not to change the Balmer test to use matrix_bb (which will separate the issue of the Balmer test from any mods to ML93 going forward.

[kslong]

I have verified that with the suggested change the Polynesia branch passes all of the github checks.

[kslong]

Polynesia branch has been merged into dev.