Change from method_directory -> analysis_directory
This is to be consistent with other variable names in the script.
Adding default option set to false for saving the subset list.
Not saving is preferred for quick analysis as it will load in a file already
on disk if it exists which would mean exploration analysis is hindered.
Alternative method for sorting subset file
Previously with dtype='<U9', the object names where being clipped to be
under 9 characters. This would cause KeyErrors later in the pipeline as
the object names were changes.
The changes here should allow for a more general approach.
This has been tests on SPCC and PLASTICC datasets
REF:
https://stackoverflow.com/a/16946553/4521950