sacc files and tracers for CrossCorrelation Likelihood

[itrharrison]

This PR will change a lot about how the CCL-based CrossCorrelation likelihood will work.

So far I have added the ability to read sacc files and use the binning specified there to the ShearKappaLikelihood subclass.

[codecov-commenter]

Codecov Report

Merging #58 (7023a02) into master (5351ca2) will decrease coverage by 1.32%.
The diff coverage is 29.16%.

❗ Current head 7023a02 differs from pull request most recent head 238d9a6. Consider uploading reports for the commit 238d9a6 to get more accurate results

@@            Coverage Diff             @@
##           master      #58      +/-   ##
==========================================
- Coverage   67.34%   66.01%   -1.33%     
==========================================
  Files          29       29              
  Lines        1666     1701      +35     
==========================================
+ Hits         1122     1123       +1     
- Misses        544      578      +34     

Impacted Files	Coverage Δ
soliket/cross_correlation.py	56.41% <29.16%> (-43.59%)	⬇️

[itrharrison]

I just updated the test for the ShearKappaLikelihood which is now against data digitised from Figure 3. of https://arxiv.org/abs/1401.6018 with n(z) from Equation 8.

The direct agreement is between unbinned theory curves here:

whilst the test is done with a binned version of the data, which is in the data file and has agreement shown here:

[itrharrison]

As an aside, I am maybe uncomfortable with the fact that sacc_data.get_bandpower_windows(bpw_idx).values was able to return floats which then cause ccl.cls.angular_cl to return zero, so I just cast them to ints.

I don't know if @damonge thinks this is worth a sacc (and/or CCL) Issue?

[Pablo-Lemos]

This looks really good, thanks for putting the work!