See ews-epidemiology commit 922332e.

This creates input files for an epi run that are the output of an epi run from the previous day. Connects with ews-epidemiology development at the same time.

Remaining tasks to finish this work:
1) Stop loading env and depo from days before today
2) Write tests

This is done with a set of <date_valid_from>: <host_file.tif> pairs in
the run configuration under ['Epidemiology']['Host']['HostRasters'].
This is an optional argument. The 'HostRaster' argument in the
configuration is still needed, but is renamed as 'TargetRaster' to
avoid confusion.

Cocnurrent changes were made in epi model.

This may one day be replaced by some 'host preprocessing' type
of EWS component.

Aspects to improve:

*  Currently, the dynamic host map has to be handled both in
ews-epidmiology for run time, and here for post-run in order
to convert units. It might be neater to do once in one place only
 by moving the unit conversion functionality into the epi model.

 * Implement testing for handling of dynamic host data.

See EpiModel.EpiPrep.updater.update_input() docstring for comments.

This should substantially speed up production epi runs, particularly
towards the end of the season.

This commit also unifies deposition and env suit preparation code.

So that coordinator tests work with ews-epidemiology commit d344567 .

See EpiModel commit 728921c.

This is required for the standard epi model output, which has units of
ha_infected per ha_cell, because the model risk thresholds are fitted
for units of ha_infected per ha_wheat.

Optional config argument to delete mid-process files at the end of a
successful job. There is an associated command-line flag '--clearup'.

The reason for this feature is to provide a way to slim down data
volumes of production runs. A lot of processing files are generated
and production runs may only need to retain configuration, logs and
outputs. Also, the inputs might be archived separately, so they
wouldn't need to be retained in prod runs.

Currently, files are moved to a hard-coded 'recycling bin' directory,
and there is a minimal set of files set for clearup: only the deposition
and environment tar files.

This change requires some standard structure to outputs of
processor_in_job functions, which would be more robust if
a class structure is implemented.

TODO: Implement a test for file clearup.

Because of changes to EpiPrepLister in the last few months, the file listers have changed name and are too complicated to guess. Now, it will be necessary for the user to specify them in the input config file. EpiPrepLister's docstring will have to guide the user clearly enough.

wiring in the newly separated post_processing module into coordinator, rather than the plotting module

1) There is now a python alternative to the R-based wheat-source-
generation, called source_gen. The functionality to call source_gen
has been added (see ProcessorSurveys.pyL939 and L1032-1040), but not
yet switched on while we await testing of
output at the Met Office.

2) ProcessorComponents was huge (2100 lines!) and the structure was
not apparent. Each component's supporting functions are now in a
separate directory. There is also a ProcessorServer.py relating to
file down/up-loads. ProcessorComponents is now essentially a directory
 to import all the functions. This might inspire a more systematic
 (object-oriented?) structure for each component within coordinator
 in the future.