test whether we can run NUTS without warmup on different noise realizations of same galaxy

[ismael-mendoza]

Using the median tuned_parameters from a sample of ~100 noise realizationns. This could enable us to run the shear inference test much faster, concretley if no warmup, 100 samples per noise realization takes ~0.1 s ==> 1e5 galaxies ==> 1000 s ~ 16 min

• check efficiency of samples
• check r-hat
• check a couple of trace plots