updated DFT_QCI_thermo with uncertainty data from paper

[nickvandewiele]

also: updated indeces > 33 to comply with original indices of paper.

[nickvandewiele]

This pull request solves ReactionMechanismGenerator/RMG-Py#267

The script to do this can be re-used:
https://gist.github.com/d6823c65e8a6b16debc1

[rwest]

Looks good, thanks for doing this.
And I think the Travis failure is because of the new adjacency list parser not noticing the difference between O_atom_triplet and O_atom_singlet and is unrelated to this.

However, I do wonder what we should do about the uncertainties that are reported as zero. Presumably @cfgold calculated that they were less than 0.05 and he was reporting things to 1 decimal place so rounded to 0.0 instead of 0.1. However, if one day this gets used in some optimization or "consistent set" checking tool (as we saw many of at the Symposium), then an uncertainty of precisely zero will make the constraints hard to satisfy! Should we put 0.05?

[nickvandewiele]

Agreed. I will update all 0.0s to 0.05s.