issue with error determination in the FastFit

[mmosc]

Hi David,
I think there is a problem in the determination of the errors for the FastFit.
You solve:

err_l = brentq(lambda x: -2*fit.log_likelihood([x]) - 1, a=0.03, b=bf['x'])

err_r = brentq(lambda x: -2*fit.log_likelihood([x]) - 1, a=bf['x'], b=0.05)

but we should actually solve Delta Chi^2 == 0, i.e.:

err_l = brentq(lambda x: -2*(fit.log_likelihood([x])-bf['log_likelihood']) - 1 , a=0.03, b=bf['x'])
err_r = brentq(lambda x: -2*(fit.log_likelihood([x])-bf['log_likelihood']) - 1 , a=bf['x'], ##b=0.05)
Best,
Marta