Allow U-Map to run in parallel to increase performance

[stefanhahmann]

When performing dimensionality reduction using U-Map (cf. e.g. https://www.youtube.com/watch?v=qZ8KDrgL1Ro) the computation can take very long to large (>100.000 points) datasets (>5 minutes).

The reason is that the U-Map is not run in parallel mode:

While this makes sure that results are reproducible, it would be nice, to have a checkbox that allows the user to between slower performance and reproducible results and quicker performance and not exactly reproducible (but similar) results.

[haesleinhuepf]

Hint: we deactivated parallel execution because it crashed when called from Jupyter:

• https://github.com/BiAPoL/napari-clusters-plotter/issues/169

Maybe it's possible to detect if we run from within napari and then enable parallel execution.

[stefanhahmann]

Thanks for the advice. I did not see the optional checkbox to enable multithreading when running napari from command line. But now I found it.