- allow likelihood to use a rate function with a
   dependent variable
 - refs Monash/tupak!132

 - refs Monash/tupak!132

Resolve "`sample_subset` in `PriorSet` should not have have `list()` as a default argument for `keys`"

Previously, outdir was checked and if-not-existing created in a few
different places in two different ways. This (a) makes them all use the
same method (namely the utils function) and (b) places a
check-and-creation call before every tupak output. This will ensure that
changes elsewhere in the code won't cause bugs in the future. Also, if
you choose not to save anything, it won't just make an empty directory.

1. Removes the samples - these are duplicates of the posterior data
frame
2. Adds option to save the results. For large MC runs, it's not always
neccersery to save the output.
3.

1. Explicit with/without check pointing functions
2. Closes #150 if timing is not succesful it defaults to no
checkpointing
3. General clean up of some doc strings
4. General clean up and commenting of kwargs setup

This was rarely used and should instead be implemented as another
sampler.

keys() now returns a dict_keys object which is a view. So when the
actual keys are updated, so is this, but we do not want that behaviour.

1) If sigma is not given, then it will not be sampled over (or a warning
raised that you didn't specify a prior for sigma).
2) Comments added to explain the logic throughout
3) Tests added for the Gaussian likelihood

Previously, the frequency_array was passed to infft, but it should
instead be the sampling frequency.

Implements automated burn-in cuts for emcee