Previously, each search class named its own output file. This provides a
function `set_out_file` in the top-level class which sets the `out_file`
attribute to be used.
- Adds class name to the out_file for clarity
- Add detectors used

If there are multiple detectors, then multiple .sft files are generated.
This updates the attribute 'sftfilename' -> 'sftfilenames', a list of
those files while also updating 'sftpilpath' to a colon separated list
of the relevant sft files.

- In a prior commit, the number of sidebands checked for the normal search
(i.e. SSBprec=2) was erroneously set to M0, rather than 9. This fixes
that.
- Also improves logging output
- Splits of the UniformPriorToGrid run function to produce the data and
  plot separately

The command line argument is matched to formatting used by MFD to
reconstruct the argument (using .16g) formatting. Ideally, all REAL8
should use this formatting.

Removes the forced `int` for `fmin` in MFD. As noted by David, this
isn't required.

 -ported check from MCMC search
 -bug: for single-str SFT path, glob list comprehension
  was trying every single character

 -try-except wrapper for subprocess.check_output()
 -use this for MFDv5 call (properly terminate at error!)
 -and for other small utility calls
 -use unique variable names for cl strings too

 -first check $DISPLAY environment variable
 -then use non-interactive "AGG" backend if necessary

- Adds ability to use either a dict or a par file
- Fixes bug in which if tstart and tend aren't specificed the data
  saving mechanism breaks

- Add peak detection utilities
- Move generate_loudest
- Add SSBprec pass through to MCMC searches
- Fix convergence stat (missing sqrt)
- Minor fixes to plot_walkers (put detection statistic behind walkers)

- Adds injectSources to a variety of other methods
- Adds conversion of injectSource to pfs_input

- Uses the actual prior and not the log-prior as previous
- Adds option to plot the full extent (currently only for uniform
  priors)
- ensures the xlims and ylims are reset correctly

If the given tstart is before the minStartTime this will error - so
instead, select the minStartTime

Previously, if the data span was shorter than the requested tstart-tend,
the code would choke when attempting to calculate twoF over a range
without any data. Instead, the code now automatically only compute twoF
over a range with data in it.

CI was failing when calling the new injection_helper_functions module as
anaconda attempted to update numpy (unknown reason)

1) Add ability to in F0s, F1s etc in grid search directly (use tuple as (is)
2) Update cumulative legend
3) Add option to produce pfs_input directly to calculate_pfs
4) Fix spelling mistake
5) Update transient example in the paper

1) Adds check if the truths are not of the correct length (to prevent
corner.corner error)
2) Improve sliding window labels
3) Adds option to plot gps seconds on the sliding window

A search, based on Reinhard's J0537-6910 work, in which the search is
performed over F0 in a sliding window. Able to track changes in the
power and frequency of the signal.

- Adds option to underplot the preducted Fstat 1-sigma variation in
  twoF_cumulative plots
- Adds isolated function to predict Fstat (wrapper to lalapps function)
- Adds detector colours dictionary: use elsewhere

- Expose generate_loudest
- add check if par exists
- All read_par to read_loudest
- Make the default save_fig to be True in plot_corner