Pull ghcrawler via submodules rather than individual clones

[joedborg]

Makes the install procedure a little more smooth by using submodules, rather than individual clones of dependencies.

[stuartlangridge]

I'm not sure about this. I grant it makes it easier to pull and develop on the whole lot, but makedash doesn't actually need to be run in the same place as ghcrawler, and it's quite useful to keep that separation. For production deployment, the long-term plan is to provide a docker image, probably (this is still under discussion, and thoughts are welcomed). I don't really want to tie pulling all the ghcrawler stuff to pulling Measure itself. @jeremy-lq do you have thoughts?

[joedborg]

Yeah i totally agree re. "proper" deployment. My thought there would be that you just don't then init the submodules, but you're definitely still pulling more than needed in that case (the submodule definitions). It's just that for someone spinning it up locally to try out, it's less things to worry about. Also, the possible need to pin ghcrawler to a commit if needed in the future.

Once I've got a stack running and collecting data locally, that aim is to get it running in Kubernetes (the manifests of which, I'll obviously share).

[jeremy-lq]

Given the points made by @stuartlangridge along with the fact that we need to figure our a longer term strategy for supporting sources beyond GitHub, it may make sense to hold off on merging this.