Added basic inequality mesasures: lorenz curve and gini

[cdagnino]

First stab at creating inequality indices, as discussed in #395

Tests added and nosetests passed.

[coveralls]

Coverage decreased (-0.8%) to 94.405% when pulling 176801b on cdagnino:inequality_measures into 8657c91 on QuantEcon:master.

[jstac]

Thanks @cdagnino

For me, prange is not working once you wrap computation of the Gini coefficient in a jitclass. I can see that by monitoring system resources.

Since jitclass is still experimental and seems to conflict with prange, I recommend that we keep computation of the Gini coefficient as a function with @njit(parallel=True) as a decorator.

I can see your motivation for wrapping the functionality in a class, since you propose additional methods, but the cost of preserving these operations as individual functions is not high.

[mmcky]

thanks @cdagnino this is looking good. I agree with @jstac comment re jitclass.

In other code I often write jit functions that do a lot of heavy lifting which makes them accessible (through import) at the function level, and when there is a use case I often use those fast functions to construct a convenience class that holds data and methods and in turn make calls the collection of general functions.

[jstac]

@cdagnino Thanks for your help to date. Should we leave this to you or do you not have time?

[cdagnino]

@jstac @mmcky Apologies for the (really) long silence. I've pushed the changes you suggested. Let me know if I should do a pull request now

[rebekahanne]

@cdagnino This looks good. I just made a few comments that might help the documentation be a bit cleaner.

[mmcky]

thanks @cdagnino and @rebekahanne -- look forward to merging this.

[cdagnino]

@mmcky @rebekahanne I pushed the suggested changes to the docs. Let me know if anything else is missing :-D

[mmcky]

thanks @cdagnino and @rebekahanne. I have tagged this as ready and it will be merged later today.