Moritz Huebner: Introduced a __repr__format_helper method to help automating __repr__ formatting for all subclasses.

Makes the decoding less aggressive. This was causing various things to
be coerced into a list when they where dictionaries etc.

- Reduces the length of the output
- Explicitly states if it is a logz ratio or just logz

If lalsim is not installed just add a warning rather than raising an
error. Previously this did already exist in source.py, but we now also
import lalsim in conversion.py as well

- Renames run_sampler to run_external_sampler to differentiate it from
  the tupak run_sampler command
- Minor fixes elsewhere to code

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

- Renames the functin to show it is the `posterior` attribute which is
  generated

- Add the label to the saved prior
- Remove obsolete write function

Fixes one issue in #75

Previously renamed example not updated in tests

min and max as properties for interpolated prior

See merge request Monash/tupak!47

Moritz Huebner: Added check if the prior has been set to None. Will set it to an empty dict in that case.

Adds decoding for single strings and list of strings

See merge request Monash/tupak!50

Fix one of the sub issues in #75. Namely plot_corner will now run for
data generated by python 2 when imported into python 3

An issue with html5lib upstream is causing the docs to not be generated.
This ensures a version is used that will make the docs.