Update routines to call OneCovariance for tomography

[sachaguer]

This PR should be quite small. OneCovariance already handles tomographic run. These commits are aimed at modifying the functions setting up the config file to compute a tomographic Cell covariance with OneCovariance.

In particular, it requires the computation of shape noise and galaxy density per tomographic bin.

This PR is related to the issue #283 .

[cailmdaley]

Nice and compact — the config plumbing does what it says. Two small initialization gaps will KeyError on first use; both are one-line fixes.

— Claude, on behalf of Cail

[cailmdaley]

oops sorry Sacha! i meant to review CosmoStat/shear_psf_leakage#35 instead of this PR, which i see may be in progress and which you hadn't requested my review on yet.

[sachaguer]

The OneCovariance wrapper now runs in tomographic case (see figure below).

@cailmdaley this is ready for a review. A few things to note during the review:

• I checked, but worth keeping an eye on it during the review, that the ordering of the blocks in the covariance match between the iNKA merge I wrote and the way OneCovariance saves the data natively. I looked at the diagonal of the two covariances and there is a decent match so I think this is correctly done.
• I met some hurdles on the way related to package versioning. We currently install numpy==2.5.0 (the most recent version?) using the pyproject.yoml. But OneCovariance crashes with numpy>2. I opened an issue Runtime error when using numpy >= 2 rreischke/OneCovariance#34. I updated the code to run on my own fork and OneCovariance is in my knowledge not part of the package yet. We might want to add it reading my fork for now but potentially from Robert Reishke's repository if he solves the issue.
• I also made some updates to the iNKA part.