Split cluster count covariance into SSC and Gaussian builders.

[mattkwiecien]

Quick follow up to the previous PR and addressing issue #65

• Split the one covariance builder for cluster counts into ClusterCountsGaussian and ClusterCountsSSC.
• Each one performs the relevant calculation and additionally returns the covariance for the block as an array of a single float, rather than just a float, to work better with the existing code.
• Updated tests and notebook example.

[mattkwiecien]

In the notebook, you could show the Gaussian and SSC terms independently too.

Good idea, will add.

Where is tests/data/clusters/mock_clusters_with_cov.sacc used? Wasn't it there before?

I think it makes sense to have 2 sacc files here. One that has no covariance, that a user can use to calculate the covariance from scratch. Then a 2nd that has the covariance already calculated and saved, so a user doesn't need to run a 1hr + notebook cell in order to work with the covariance in the notebook. If you don't think we need that I can delete it, just let me know.

In the tests, how can it be that you have more tests that are for the Gaussian than for the SSC. I would expect that SSC will need more involved methods that computing the shot-noise.

I'm testing methods in the base class, not methods in either concrete implementation. So I just need to instantiate one of the classes in order to obtain methods in the base class. So it's not really testing the gaussian implementation but the shared methods.