Correct log-likelihood docstrings and make them consistent

[k-shep]

The Gaussian log-likelihood description (https://pints.readthedocs.io/en/stable/log_likelihoods.html#pints.GaussianLogLikelihood) in the MultiOutputProblem case is incorrect. There should be i dependence in the final sum.

The log-likelihood descriptions are also inconsistent between different log-likelihoods. For instance, the MultiplicativeGaussianLogLikelihood (https://pints.readthedocs.io/en/stable/log_likelihoods.html#pints.MultiplicativeGaussianLogLikelihood) has the i, j in different orders compared to the Gaussian log-likelihood documentation.