Adjust cluster covariance calculation to use new SACC file tracers

[mattkwiecien]

There's been development on the cluster pipeline that has led to improvements for how we store cluster data points in sacc. Each data point for a cluster data type comes with a (survey, z, richness) tracer combination. This PR updates the code to extract the data from that new sacc file format.

Closes #91

• Requiring sacc >= 0.9. This includes the new cluster changes and has no breaking changes.
• Removing old sacc files
• Updating the example notebook and checking in a new mock sacc file.
• Modified load_from_sacc to use the new data structure. Only significant change here is now richnesses are being saved into sacc in log base 10. I updated the code to read this in so the existing behavior doesnt change
• CovarianceClusterCounts is now overriding _build_matrix_from_blocks and has been renamed. The reason is because before the tracers were unique per data point. This is not the case (we have the same tracers for cluster mass and cluster counts). We need to therefore specify the data_type we care about when building the final covariance matrix, otherwise we will populate other parts of the covariance matrix (e.g. cl mass) with the cl counts covariance.
• Added CovarianceClusterMass covariance calculator as a placeholder for the cluster weak lensing mass covariance matrix. Currently this is not implemented anywhere and needs to be established. For now it's simply the standard deviation of the masses in a given bin. By the time TJPCov is reading the sacc file this data is just a single point, so for now TJPCov just reads the previously calculated diagonal covariance and returns the same value. This will be changed in the future.

[mattkwiecien]

Updated the summary - added a new CovarianceClusterMass to hold the diagonal covariance matrix for cluster mass errors. This will be changed in the future but at least for now we are returning something for cluster mass covariance.

[mattkwiecien]

I thought google required single sentence not single line, my bad, fixed!

[mattkwiecien]

Merging without tests passing because this is the MPI bug addressed in #93