Previously I was setting the noise_log_likelihood to nan in a confusing
way. This removes that and also renames all the liklihoods for clarity.

The injectino parameters do not correspond to the recovered parameters
(two different models).

and time marg before moving onto dist and tiem nad phase marg

- Add doc file on CBC PE
- Add introduction text

ratio looks a bit off: need to figuring out what's going on with the
normalization

Allows easier access to the Likelihood, GWT, WaveformGenerator, and
run_sampler

- Fixes bug when setting truths
- Remove mu=None, sigma=None in hyper PE example

- Adds a tests flag for which the sampler only runs 100 steps
- Set the samples as random samples to test plotting
- Fix problems in a few of the examples