Integrate cross-project experiment into pipeline

[salsolatragus]

No description provided.

[M8is]

In crossproject_prepare.py, line 36, we open a file for reading for which I can't find where that file would have been written before. I've searched the repository and this is the only usage I found.
@salsolatragus Do you know how subtypes.csv would be created?

[salsolatragus]

I created this file manually, but it turned out the experiment didn't actually use it, due do us limiting the number of examples projects to 50 per type. For now, I suggest you assume this file is empty (or non-existent). This is how I used this code in the last submission.

[salsolatragus]

• This is more a reminder for myself: I'm not happy with the "create project list" part. Currently, it always creates a list after prepare, considering only the projects relevant to the current run. What I actually need is a list with all the projects relevant for the experiments we've run. Of course, I can rerun the preparation with respective filters, which will generate the desired list in the end, but this doesn't really feel like a good solution... on the other hand, I don't really have a better idea from the top of my head :-/