using caching for get_runs_all_subjects to speed up get_runs

[agramfort]

closes #715

@apmellot can you see if it fixes your problem?

[apmellot]

@agramfort Yes it fixes my problem!

[hoechenberger]

I didn't really take the time to look, but I expected the pkg_resources issue to be solved already upstream via mne-tools/mne-python#11454 -- maybe some bits were forgotten there?

[larsoner]

@hoechenberger downloading eeg_matchingpennies is currently at ~40 minutes and will probably time out soon. It's a ~350 MB dataset for subject 05 (the one we use) -- I'm going to just make an osf.io mirror of the dataset I think...

[agramfort]

I just restarted the CI

…

Message ID: ***@***.***>

[larsoner]

I just restarted the CI

I've effectively done that a couple of times, and it didn't help. I am downloading the dataset now and it has been over 20 minutes and it's not done locally. We need to migrate I think

[agramfort]

sounds good!

…

Message ID: ***@***.***>

[hoechenberger]

I'm going to just make an osf.io mirror of the dataset I think...

Go for it :)