epsilon is not properly propagated in computation of VRMSE

[gcouairon]

Hi,
I see that the epsilon used for validation in the VRMSE metric is not properly propagated to NRMSE.

the_well/the_well/benchmark/metrics/spatial.py

Line 240 in 6cd3c44

return NRMSE.eval(x, y, meta, norm_mode="std")

Does that mean that the validation epsilon used in Walrus is actually 1e-7 and not 1e-5 ? defined here).

Thank you !