Circular pipeline for stages that require theory inputs

[elvinpoole]

We currently can't use the theory data vector as an input to any stages that feed into the data vector measurement as this will create a circular pipeline. This is because the theory stages require the measured data vector SACC file as input, mostly for metadata (and n(z)?).

e.g. if TXLSSWeights needs to use a theory w(theta) for sample variance estimates:

flowchart LR
  A[TXLSSWeights] --> B((binned_lens_catalog))
  B --> C[TXTwoPoint]
  C --> D[TXTwoPointTheoryReal]
  D --> A

Loading

Possible options
(1) Generate a new theory vector within each stage that needs one. Might requires some tweaks to utils.theory. Need to keep track of all the data vectors (add it as an output to these stages?)
(2) Remove twopoint_data as an input to the theory stage. Theory stage might need to be run multiple times. e.g. with different theta binning