Add hybrid modeling with neural-network-based binding models

[jbreue16]

This PR introduces hybrid models to CADET-Core, allowing the binding model to be replaced or approximated by a pre-trained neural network.

Todo

• get rid of src/mkl stuff and use Eigen3 instead for matrix vector operations
• rebase code
  • resolve conflicts
  • add changes to generalized unit operations ColumnModel1D.cpp, ColumnModel2D.cpp
  • squash commits, are there meaningful separate commits?
• add documentation
• add more cadet-core tests?
• add simple python code example to cadet-verification, where a NN is trained and plugged in
• Give some h5 file references to be incorporated into cadet-verification by @jbreue16 , preferably settings used in publications

[github-actions]

Thank you for your submission, we really appreciate it!

Like many open-source projects we ask that you sign our Contributor License Agreement before we can accept your contribution.
To sign, please post two separate comments based on the following templates 👇

1. Comment:

---

- [ ] The CADET maintainers know my real name.

At least one of the following two applies:

- [ ] The CADET maintainers know my current employer.

- [ ] I am *not* making this contribution on behalf of my current employer.

---

2. Comment:

   ---

   I have read the CLA Document and I hereby sign the CLA.

   ---

   _{You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.}

[schmoelder]

Unfortunately, rebasing was not possible without additional work. Amongst other issues, the PR includes changes to the unit operation models and it's not quite clear yet as to why. I see the following options:

• Investigate what other changes have been made in this PR. Maybe we're lucky and we figure it out quickly, but it could also be tedious
• Start from scratch, using this PR merely as inspiration

Any preference @lieres?

[lieres]

@jazib-hassan-juelich, please have a look and comment.