foxtran

I think I can use detrotra8 procedure to set rot/tra freqs properly.

Still can be reproduced with latest xtb version. I will fix it at morning.

Feel free to fix bugs. This is **necessary** and **urgent** for the next release. See issue #1134 for the blocking issues. Or just use main branch like I do :-)

> For builds from source however it would be great to include the latest version. See #1120

I think one needs to setup something like dependabot

Commits https://github.com/grimme-lab/xtb/commit/fc512bbc2528d03406c9204ddc9c03032433c390 and https://github.com/grimme-lab/xtb/commit/df80c8b3618eb8e1a47d01836d11077cd8c203b4 can give this difference. Especially, you have some coords which are equals for some axes.

https://github.com/grimme-lab/xtb/commit/65e9c14cd387f70bae4ecd369e878a4d71f19925 This commit changes how approximate hessian is computed. https://github.com/grimme-lab/xtb/commit/10c8a31cd6c9e1ff7a8ecf4db6d9666f39b24215 And here some fixes were in optimizer (but I do not see how can it affect your example)

There is an archive with some structures from 30 to 6000 atoms. 0090_0030xWATER.xyz was taken from https://ravel.pctc.uni-kiel.de/research-topics/ttm2-f-water-clusters/ 0417_0139xWATER.xyz 0570_0190xWATER.xyz, 0711_0237xWATER.xyz, 0909_0303xWATER.xyz, 1707_0569xWATER.xyz were taken from http://www.ergoscf.org/xyz/h2o.php GluAla.xyz's were taken from...

Ok. I will leave only source code changes, and return tomorrow with another PR :)

What is about #1168 ?