refactoring reading of list like files to support preserving index col type where it might be mixed -- and casting to int where needed.
refactoring other pstfrom methods to preserve model file datatypes under numpy 2 -- removing reliance on sidx (which wasn't stored well in csv under numpy2) to align pars during the apply methods. Instead use index_col values directly.
updated tests
np.NaN changed to np.nan
Fixing pstfrom add_pars to make better used of x and y or i and j in index cols.
coverage: 79.412% (+0.07%) from 79.346%
when pulling 88dacc3a23be18f620eca260a200dff2372560e6 on briochh:hotfix_numpy2
into 5551e161bb4978c599b7352ee373b0d88d831d68 on pypest:numpy2.