Lesson 15 demo update

Made some updates to the Lesson 15: data wrangling demo notebook. Switched to the feature used for the outlier analysis section to MedInc rather than MedHouseVal - I think it gives more interesting and informative results. You can easily re-run the notebook with any feature you want to try it out. Just update the missing_feature and/or outlier_feature variables in the setup section at the start of the notebook. Also switched to using Scikit-learn’s KBinsDiscretizer rather than pd.cut() for outlier binning. KBinsDiscretizer uses quantile binning - this is what we want, not equal size bins.