With low resampling rate it can happen that a template detects itself with
a very low correlation-coefficient. This patch avoid the creation of an
additional template

Update plotTimeline.sh to also plot gaps in the "scientific" theme if a gap-file "gaps_scan.dat" is given in the results folder. The gap-file is in the format of obspy_scan's --print-gaps optipon. 

(for instance, it avoids inserting catalog events into the combo catalog if
they were actually detected)

(switch b/w transfer-function-based & sos-based filtering)

(i.e. better deal with data masks of Stream/Trace objects)

... as presented in our paper as a post-processing step to reduce the weight of
impulsive waveform features on the cross-correlation coefficient

(but this doesn't include the machine learning itself... this will be provided in due time)

+ some tweaks of the old approach

(particularly to make the Basel script run smooth)

+ set name of the tool to TMT
+ improve descriptions of modules and scripts
+ clean up files in `scripts` dir

New post-processing command line argument plot-FMD
+ overhaul whole FMD plotting routine
+ outsource FMD parameter value estimation into a class
+ helper class to read/write FMD parameter value estimates to/from a json file
   (useful to reload bootstrap estimates based on a large number of iterations
    --> no need to repeat bootstrapping when changing only the plot binning,
          for instance, while the events are the same and the configuration is similar)

Supported Mc estimation methods: MAXC, MBASS, MBS, FIX.
These and more settings can be set up in the last section of `postprocessing.py`
(e.g., the time range, the binning for Mc estimation, the Mc correction factor,
the plot binning, and the bootstrap iterations).

 - wrong backwards compatibility of convolve <-> correlate
 - TM.io.write_combined_catalog: facilitate loading det files with threshold suffix
 - TM.eventops.combine_catalogs: report which events got filled in
 - TM.io.get_detection_waveforms: extract events with correct timing/length

+ also renames the 'convolve' string internally to 'correlate'
  (but is still backwards compatible)

Addresses (original) issue #6. New flag: 'X' ("no data"). Like '/', but these events
will not (tried to) be preloaded (but included in the final catalog). 

For now, this flag needs to be manually specified.
--> TODO: set this flag automatically (issue update)

(update for commit ff24946d, fa76fd82, f5e85d6d)

Previously, some magnitudes in the `regression_catalog.dat` could be wrong
if these events were excluded from the regression itself.
Now, the correct magnitudes are extracted simply from the detection list.

- remove statsmodels dependency

+ improve get_amplitude docstring

 (in particular the diagonal lines in the slope-removed plot)

+ some better formatting
+ only plot confidence intervals if number of clusters <= 3

see https://github.com/scipy/scipy/pull/10076
(sometimes, scipy.optimize.curve_fit() failed if the dtypes differ)

Write out the detection list with filled-in catalog events that are missing (i.e. that
were not detected). Can be achieved with a new command line argument:
    --post-processing write-full-catalog

This creates a new file: `detection-catalog-combo_Mag_thresh0.50.dat`

(update for commit ff24946d & fa76fd82)

In particular: non-detected catalog events are excluded from regressions and
their magnitude is kept (for `regression_catalog.dat`)

+ better logging + doc
+ fixed cluster grouping when cluster '-1' present

(update for commit ff24946d)

 Now: properly handle filtered catalog events (i.e. within desired mag limit)
      that are associated with template events OUT of mag limit
      They can form a cluster/family, which is used for regression and applied
      to the detections.
      E.g., Template 3 is out of mag limit, but it detected several catalog
      events that are in the mag limit. Then, the regression for this cluster
      is applied to all detections found with template 3.
 
 + various minor tweaks, bug fixes, & better logging

+ providing a necessary function for previous commit 47802e0b

+ allow to fix the slope of the regression(s)
+ allow to plot Amp-Mag regression lines with the slope removed --> horizontal
+ refactoring & revision of related code
+ fixes a bug of 11050b3d (excl. catalog events didn't show up in the catalog)

The config file has now two new parameters:
    MAG_REGRESS_TYPE: specifies whether to do the (old) all-in-one the or (new)
                      cluster-based/template-family-based regression)
    FIX_AMP_MAG_SLOPE: if provided, the slope of the regression(s) can be fixed

In case of a cluster-based / template-family-based regression, the cluster
association will be extracted from the detection list (i.e. template
associations).

Other:
    - optional: account for magnitude uncertainties in the regression
      (via weighting)
    - outsourced plotting of confidence and prediction intervals
      (TM.plotting._plot_CI_PI)

Notes:
    - regression plots do not have the '_w1' suffix anymore
    - renamed `regression_catalog_used-maglim.dat` to
              `regression_catalog_used.dat`

New in save_iter_to_file():
 - the callable function is optional
   (which converts an iterable's list element to dict)
    --> no need to specify custom function if one only needs to
        dump the iterable's data using the `columns`'s names.
 - write out padded column names (to correspond with column widths)

New in load_iter_from_file():
 - the callable function is optional
   (which converts the file's rows (as dict) to a list)
   --> reads all the data as strings
 - strip whitespace of column names to allow padded column names

New: started to use numpydoc's docstring format:
https://numpydoc.readthedocs.io/en/latest/format.html

... using scipy.signal.choose_conv_method()
(only for scipy >= 0.19)

Only required if SAVE_DETECTIONS = NO in the config file
   (which allows to only extract the waveforms for detections
    that end up in the uniqued detection list)

How to use/call:
    python3 xcorr.py <config> post-processing get-detection-waveforms