Dsigma

Added
- added `dsigma.stacking.mean_critical_surface_density`

Changed
- photometric redshift correction now always applied when computing the mean source redshift
- `dsigma.stacking.lens_magnification_bias` now uses `dsigma.stacking.mean_critical_surface_density` to estimate the critical surface density and not calculate it based on the mean lens and source redshift
- `dsigma.physics.lens_magnification_shear_bias` can now use angles expressed with `astropy` units

Added

- `dsigma.stacking.lens_magnification_bias` can now be used to compute the bias in the tangential shear

Changed

- mean source redshift now takes into account n(z)'s passed to `dsigma.precompute.precompute`

Fixed

- incompatibility with numpy 1.24
- bug in `dsigma.stacking.tangential_shear` when `random_subtraction=True`
- error in tomographic redshift bin assignment for KiDS, sources with photo-z's at the bin edges were assigned to the wrong tomographic bin, this biased KiDS lensing measurements by order 2%

Changed

- `dsigma.jackknife.compress_jackknife_fields` now suppresses `numpy` warnings if columns contain NaNs

Fixed

- bug in the calculation of the photo-z dilution correction factor, led to percent-level biases in the total galaxy-galaxy lensing amplitude, did not affect DES and KiDS calculations since those are based on n(z)'s, bug was introduced in version 0.6

Changed

- `dsigma.precompute.add_precompute_results` has been renamed to `dsigma.precompute.precompute`
- `dsigma.precompute.add_maximum_lens_redshift` has been removed and the functionality integrated into `dsigma.precompute.precompute` using the `lens_source_cut` argument
- `dsigma.jackknife.add_continous_fields`, `dsigma.jackknife.transfer_continous_fields`, `dsigma.jackknife.jackknife_field_centers`, and `dsigma.jackknife.add_jackknife_fields` have been merged into a single function, `dsigma.jackknife.compute_jackknife_fields`
- the computation of continuous fields to construct jackknife patches now uses DBSCAN instead of agglomerative clustering, points are also not downsampled anymore

Removed

- `dsigma.stacking.shape_noise_error`, please use jackknife resampling to estimate errors

Changed

- significant performance improvements for `dsigma.precompute.add_precompute_results`

Fixed

- crashes in `dsigma.stacking.shape_noise_error`