While working on issue#80, and PR184, have noticed that processing
config defaults to estimator.engine = "RME_RR".  This is fine, but
I find I need to specify to use "RME" explicitly when there is only one
station.  So a couple fixes were added:
1. Processing class now has a validate() method.  If there is no RR
station, _and_ the estimator.engine is "RME_RR", it gets reset to
"RME".  Also added the ability to pass a kwarg to ConfigCreator instance
called estimator.  The kwarg is a dict and if "engine" is a key, it will
overwrite the estimator with the corresponding value.

The parkfield SS run test was updated to use the config_creator method.
cas04 test is usign validate()

[Issue(s): #80]

Tested clock zero works when it comes from data, but only on a single
run.

[Issue(s): #42]

-Added windowing_scheme as a property of decimation_level metadata object, and r
eplaced initializations of WindowingScheme() in time_series_helpers, with this p
roperty
-cleaned up an errant print statement and tidied some docstrings

[Issue(s): #42]

clock zero

Fixed a few docstrings, but main change was to review the math in
squared coherence calculation.  It turns out this is not as inefficient
as I had thought, but it can be done a little cleaner.  Removed an
unneeded conjugation.

[Issue(s): #78]

See notes in issue #78 about einsum motiation.
Also set show_response_curves to false, getting some matplotlib errors
in the CI tests

[Issue(s):  #78]

Fix issue 78

Using the updated method in mth5 locally (see mth5 issue #105),
am now able to process runs c and d for CAS04 as single station.

Working on getting a similar h5 built in tests/cas04

[Issue(s): #31, #80]

Allow request list to have mulitple stations and modify
channel_summary_to_make_mth5 to groupby station,run rather
than just run.  Add tests of make multistation mth5 to cas04 tests.

[Issue(s): #80]

-replace DatasetDefintion by Dataset
and import as TFKDataset,
-replace dataset_definition with tfk_dataset

[Issue(s): #80, #132]

This is just a stage commit because all tests are passing currently.
operate_aurora is not yet working.

Need to decide where to put the RunSummary wrangling.

[Issue(s): #80, #118, #132]

Add a method to KernelDataset to extract run info, looping over runs.
Also, noticed that some synthetic tests were commented out, fixed this.
Also, tidied some code in process_mth5.

[Issue(s): #181]

Multiple runs now entered into TF XML

Replaced dict with classes.  Now have a SyntheticRun and a
SyntheticStation.
This will be used to create an example synthetic case with many runs

[Issue(s): #80]

…age used

Change from timedelta.seconds to timedelta.total_seconds()
Remove run_id from sort_by, it should be only station, starttime

[Issue(s): #80]