Add script to analyze feature generation log files

[bfhealy]

This PR adds a new tools script to read slurm log files from feature generation jobs, pairing the job runtimes with the number of initial sources. When given a text file of the form "job id,runtime" from the automated slurm emails, e.g.

"25401362,16:05:28"
"25401338,11:31:10"
"25401382,1-20:20:16"
...

this code will produce a json file output that contains information relevant for predicting the runtime of future jobs, given the number of initial sources:

{"25401362": {"n_sources_start": 25190, "n_sources_end": 24443, "runtime_seconds": 57928.0, "seconds_per_source_start": 2.2996427153632393}, "25401338": {"n_sources_start": 27633, "n_sources_end": 26724, "runtime_seconds": 41470.0, "seconds_per_source_start": 1.5007418666087649}, "25401382": {"n_sources_start": 19141, "n_sources_end": 18696, "runtime_seconds": 159616.0, "seconds_per_source_start": 8.33895825714435}}

[mcoughlin]

LGTM