Add access to chi2

[afedynitch]

This update exposes the chi2 function to the public interface.

from daemonflux import Flux

flu = Flux()
print(flu.chi2())
> 0.0
print(flu.chi2({'GSF_1': 1.0, 'GSF_5': -1.0}))
> 16.788997480515302

There are a few more minor modification when use_calibration = False. First, the case does not represent the error estimate from DDM + GSF because the gradients in daemonflux have more subdivisions in energy and because gradients are derived using an effective scheme and not the full error propagation (see Sec. II.D.2). The updated (uncalibrated) covariance matrix now resembles DDM+GSF estimate a bit better but they remain different. To disable this behavior and use uncorreleated parameters (for example in a fit), pass Flux(... , uncorrelated_hadr_errors=True, ...), which will replace the hadronic part of the covariance with a unit matrix.