Implement Sortino ratio for portfolio

[dpapakyriak]

I implemented an initial attempt for the sorting ratio, as asked in issue #401. In order to achieve this, I initially implemented a function to calculate the downside deviation for all assets. Then, I made the sorting_ratio() function to return the sorting ratio for a given portfolio. This is a draft, please review and contact me via email to fix any issues and/or any additional documentation.

[robertmartin8]

I'm not merging this for now for a couple of reasons:

1. I think there are probably much more efficient vectorised ways of computing downside deviation, something along the lines of a[a<thresh].std().
2. Variable name conventions: PyPortfolioOpt uses snake_case as per PEP8

[chriss1245]

Is there any update related to this topic?
I consider this feature may be very interesting

[tschm]

@fkiraly please close. For that purpose there are packages such as quantstats or jquantstats. Minimizing the semi variance is a different beast which is not solved here. This can be done though with cvxpy.

[fkiraly]

@tschm, can you provide your reasoning why this feature does not beling in pyportfolioopt?

It seems topical to me.

Maybe there is a wider scope discussion to be had.
(of course the PR has some issues that need to be resolved but the scope discussion is perhaps to be had first)