Fixing tests that failed. Added a function to filter dicts based on value instead of key. Added M&S data to grid options rather than custom options

own source files

add option to convert dict keys to floats when loading ensemble data : this is required for matplotlib/seaborn otherwise the plots are completely messed up

added options to the Moe loader so we can change the JSON data and reload it after clearing the cache

the period distribution renormalization now gets the low and high limits from the loaded data

update_dicts() now copies the type of dict_1 for the new dict, rather than using OrderedDict (maybe merge_dicts should do this too?)

I've fixed the _boxed formatting to strip ANSI codes when calculating string lengths (so you can use ANSI colours properly)

changed the colours for the "failed systems" warning so it's absolutely obvious

wrapped the JSON ensemble output in a write_ensemble() function which automatically gzips or bzips depending on the filename extension (or this can be explictly set in grid_options['compress_ensemble'] ).

shows:
current grid status (#/# and modulo)
estimate of time remaining and time to finish
memory use (of all processes)
time per run

add methods to grid.py:

rename_grid_variable
update_grid_variable
clean

add functions to functions.py:

convfloat
datalinedict
pad_output_distribution