Add embedding visualisation to the data explorer

[mrchtr]

The data explorer is already good for inspecting the processed text snippets. However, when we look at the embeddings the vectors arrays are not so useful. To compare different embedding models it would be nice to have a TSNE visualisation.
Depending on the dataset size the calculation of the visualisation could be quite heavy. Probably we have to use a sample of the dataset or only one partition.

At the end it could look like this: