[FEAT] Added TSMixerx model

[elephaint]

This PR adds the TSMixerx model into neuralforecast (in the paper referred to as TSMixerExt), which provides support for exogenous inputs, and demonstrates its applicaiton in a Multivariate_with_TSMixer example notebook.

• Added TSMixerx and AutoTSMixerx models to NeuralForecast
• Included tests in core class
• Included performance evaluation in example notebook.
• Updated some codeparts from TSMixer for consistency with TSMixerx

[review-notebook-app[bot]]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[candalfigomoro]

If TSMixerx does support exog variables while TSMixer does not, should something like this be added to TSMixer?

        # TSMixer does not support historic exogenous variables
        if hist_exog_list is not None:
            raise Exception("TSMixer does not support historic exogenous variables.")

        # ... same for futr and static

As in models that do not support exog variables (NBEATS, PatchTST etc.)

[elephaint]

If TSMixerx does support exog variables while TSMixer does not, should something like this be added to TSMixer?

        # TSMixer does not support historic exogenous variables
        if hist_exog_list is not None:
            raise Exception("TSMixer does not support historic exogenous variables.")

        # ... same for futr and static

As in models that do not support exog variables (NBEATS, PatchTST etc.)

You're right - thanks for the suggestion. I'll make it part of the tsmixerx PR.