Cache Spago dependencies and purs output in CI

[triallax]

Fixes #342.

[sharkdp]

Thank you for looking into this. I am usually very careful with caching in CI. If we absolutely must bring down CI times, that's certainly a good idea. But I have seen too many hard-to-debug errors that appeared because cache-invalidation didn't work for some reason (see your example with compiler versions, for example).

[triallax]

That is a valid concern. Just to be clear, the compiler versions thing I mentioned was just a thing to investigate before implementing this; I do not know of any cases where it was an actual issue. Anyway, I think this comment and this one show that this should be safe. Additionally, resetting the cache or disabling this is trivial should it ever prove to be an issue.

[sharkdp]

Additionally, resetting the cache or disabling this is trivial should it ever prove to be an issue.

That's the thing. Sometimes it's hard to know (or remember) that CI problems could be related to caching.

But I'm okay with giving this a try. Thank you!