Add a Corpus Download Step

[varisd]

We should implement a CorpusStep that downloads existing URL and sets it up for the follow up steps.
This step would ideally be included in usage examples to avoid making users setting up the data directories (for RawStep) themselves.
In the end, a pipeline with the download step should be executed out-of-the-box using single init and run command without any knowledge about OpusPocus.

[bhaddow]

This is an essential step for model building, but should OpusPocus handle downloading itself, or defer it to OpusCleaner (or even mtdata?)

[varisd]

If I remember correctly, OpusCleaner did not support corpus download outside of the web UI at the time of integration to OpusPocus so downloading was avoided. However, I am not against putting this burden on OpusCleaner if it can be done by it.

On the other hand, in practice, the end user will still need to get the filter.json files for OpusCleaner cleaning from somewhere which gets us back to the original question whether to just include dedicated step which downloads data and filters.jsons from a specific URL (data na filters can be in different locations)

[bhaddow]

There is now an opuscleaner-download binary in OpusCleaner (I need to get it into PyPi).

But yes, we still need for OpusPocus to download the filter.json and call this.