Change clip value in `average_spectra` and elsewhere to be smaller.

[aewallwi]

1e40 is not a large enough upper bound clip value for calculating upper bounds on weighted error bars since we routinely take (> 1e10 mK^2 Mpc^3)^-4 when we calculate averaged error bars when inverse variance weighting and leads to confusing artifacts.

hera_pspec/hera_pspec/grouping.py

Line 394 in b3f5ae4

weighted_errws = errws[stat].clip(0, 1e40) * w**2

This issue might be coming up elsewhere and we might want to revisit the clipping across average_spectra.