1) Adds Tcoh to plots
2) prints max det stat
3) Fixes total number of stages

- Other minor changes included

This adds a class to perform follow ups from semi-coherent to
fully-coherent using an MCMC search.

- Also changes the logic of the MCMC search to check for old data only
  in the run method.
- The nsteps argument exists for the MCMCFOllowUP, but is unused.
- Some work on the plot_walkers to allow for overplotting. It may be
  useful to use this for the usual calls as well so that we don't have
  an initial and walkers plot?

1) Adds binary to transient search
2) Ignores nans when generating histogram

- Also changes the parameter vector in the normal fully coherent search
  to remove the tstart and tend. This is done by wrappin the
  run_computefstatistic_single_point in a compute_fullycoherent_det_stat
  function
- Removes theta0 from the saved data dict for the MCMCSearch
- Adds example usage of the semi-coherent search

Note - There appears to be some difference between the integration plot
and the output value. This may be related to the Riemann sum issue in
emcee itself. Eventually, we should have a case where we calculate the
Bayes factor directly and compare this.

These where potentially confusing, deprecated values for arbitrarily
modifying the BSGL value. This will break some current search codes
(most notably for A14 glitches), but it is better to do so now and keep
future code clean

1) Fixes missing passing of binary argument ot the search tool
(previouslly, the binary parameters where not searched over!)
2) Adds min/maxfraction to cumulative plot and fixed start times
3) Adds binary parameters to cumulative plot call

Also adds ability to write transient data and a search script to show
the results of such a transient search

It seems the tests have been failing, but only falling out of range by
~0.11 rather than 0.1. This increases the allowed range to 0.2

1) Adds BSGL labels to cumulative plot (if used)
2) Switches on transient and re-init before computing cumulative
3) Adds cumulative plot to MCMC search
4) Adds cumulative plot for MCMCGlitchSearch
5) Adds support for neghalfnorm in plot prior posterior

Adds functionality to place a prior constrain on BSGL minimum value.
This can help smooth MCMC searches, but it remains unclear if this and
the BSGL_PREFACTOR don't invoke something strange.

Also ads check that numdets>2 and reduces linewidth of walkers

By default, convert to nat. log.

This is not ideal, it would be better to calculate the idxs for the
keys, but I don't see an easy to way to do it at the momemt.

1) Make tqdm import allow error and use it for the grid search as well
2) Allow functionality to change the flatten method in the 2D plot