pseudo_cl.py reads the n(z) path from the config directly, ignoring `self.blind`

CosmologyValidation.get_redshift (core.py:383-410) applies the blind by rewriting the _[ABC].txt suffix on redshift_path. pseudo_cl.py never calls it and never references blind at all: lines 129 and 316 read self.cc[ver]["shear"]["redshift_path"] directly.

Consequence: with blind="B" or "C", the theory C_ell and the Gaussian/OneCovariance products are built from blind A's n(z), while pure_eb.py:117 honours the requested blind in the same run. Silent — same filenames, same shapes, no error. Harmonic space is therefore not actually blinded, and a non-A run is internally inconsistent between its config-space and harmonic-space products.

The fix is to go through self.get_redshift(ver) (or a shared path accessor) at both sites. #312 will retire this blinding vocabulary wholesale; until that lands the current behaviour is wrong rather than merely legacy.

Start with CosmologyValidation.get_redshift in core.py:383-410, then inspect the two direct redshift_path reads in pseudo_cl.py at lines 129 and 316 and compare them with pure_eb.py:117. Done means blind B or C uses the requested n(z) consistently for theory C_ell and Gaussian/OneCovariance products, rather than blind A.