Add plots for evaluation that only use likelihood

[ismael-mendoza]

We just add a notebook demonstrating how to do fisher analysis with the shear likelihood. But since we can run chains so fast in the toy case, I'm not sure about the utility of the Fisher stuff at the moment. Let's reopen issues if useful down the line if we have a specific question we want to answer with the Fisher formalism.

• Closes exp0: predictions on mean and std only from likelihood #51
• Closes try shear inference with different parametrization (|g|, beta) #49