Check all frequencies in detrotra

[foxtran]

Fixes #1203

H2 eigvals are the following:

2.9933945092181844E-015   8.5529038621230392E-015   2.6862916288779701E-014   3.4268547221618851        3.4268547221618966        40.610372376753845

[foxtran]

@marvinfriede, can we merge it?

[foxtran]

@marvinfriede, ping :-)

[thfroitzheim]

Isn't this the same issue we see in #1210. For linear water, it appears to be around 900 wavenumbers if I see it correctly. Why do we limit this at all? Shouldn't the ranked value be correct even if we check for all modes? If not, we need to introduce a check that at least 5 or 6 values are included for linear and non-linear molecules. Otherwise, it will fail again, right?

[foxtran]

Why do we limit this at all?

Now, there is no limits. All freqs will be checked.

Isn't this the same issue we see in https://github.com/grimme-lab/xtb/pull/1210.

They are different.

[foxtran]

Ah.. It always checks 5/6 freq., but only 3 can be imaginary :(

[thfroitzheim]

Are we actually sure that this is the right way of finding rotations and translations? As far as I know, one normally projects rotations and translations out so that the final hessian has 5/6 degrees of freedom less. The remaining degrees of freedom are then all vibrations. We have code to do something like this here, but we still try to assign frequencies to the full hessian. Maybe it is better to do it properly via a projection to make sure it works in all cases.

[foxtran]

It seems to me that this part was never been adjusted for such cases. So, let it be as is without significant updating like in #1210