Document threading performance from conda-forge install

[awvwgk]

The documentation should mention for the conda-forge install that xtb has better threading performance if used with the MKL backend for BLAS. In https://xtb-python.readthedocs.io/en/latest/installation.html#installation-with-conda only the basic setup is mentioned. Additionally, there should be an example for installing with MKL backend:

conda install -n xtb-python xtb-python 'libblas=*=*mkl*'

Related issues

• https://github.com/grimme-lab/xtb/issues/401