error when no gradient function is provided

[gica3618]

When a sampler is initialised without gradient function (that is, grad_logLikelihood=None), and error is thrown by the helpers.py:
NameError: name 'numerical_grad' is not defined
Indeed, numerical_grad on line 53 of helpers.py is not defined. So I believe that currently, the code does not work if no gradient function is provided.
As an example to reproduce the error, one can replace grad_logLikelihood by None on line 185 of imf_examples.py.

[JohannesBuchner]

yeah, grad_logLikelihood=None is not intended to be supported here, but was supported in the original code.

This is an odd error to receive, and a byproduct of stale code that should not be reached. Probably best would be to assert grad_logLikelihood is not None somewhere.

PRs are welcome.