Updates limit setting code

[ashleyrback]

Includes:

Bug fixes/updates:

• Updates to limit_config class. Limit setting code was returning scaled counts at limit rather than spectra.sum() counts. Changes _chi_squareds to have 3 columns: chi squared, current count and sum. Changed other class methods accordingly.
• Updates zero_nu_limit script:
  • Now uses smeared hdf5s
  • Decreases granularity in loops --> was too slow
  • Updated count values, priors
  • Saves systematic analysers (see below)
  • Plots signal vs chi squared
  • Adds verbosity option

New:

• Adds SystAnalyser class to limit_setting. A class for analysing the effect of systematics. Creates instances via LimitSetting.configure_background.
• Adds utilities module and timing utility
• Adds plot_chi_squared module with chi_squared_vs_signal method. More to follow.
• Adds dump_ndarray and load_ndarray functions to store module. These allow you to dump/load any class instance that is mainly numpy arrays, e.g. SystAnalyser class.

[ashleyrback]

@jwaterfield here are the bug fixes, updates and some new features for the limit setting code. I haven't run the unittests or added any new ones yet or checked the docs. I'll do that now and push any changes.

[jwaterfield]

@ashleyrback cheers reviewing now...

[ashleyrback]

Note: unittests currently give

ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

This will be fixed in the next commit that is pushed

[ashleyrback]

@jwaterfield pushed some updates, hopefully this responds to all comments. Let me know if there is anything else.

[jwaterfield]

Merging