Add unit tests, modularize covid19 parser for API

[flaneuse]

• Modularize pre-processing script
• add unit tests to the parser to double check and verify values. Also identify outliers for manual inspections: for instance, places which jump from 0 cases to thousands in a day. Anything _numIncrease should probably not be < ~ -10 for any given day; similarly _rolling should probably not be negative.
• Where possible, break code into reusable chunks if not already done, and structure ad hoc code like for cruises etc.

[flaneuse]

note from @gkarthik : "_numIncrease is negative for countries like Spain. JHU didn’t bother to correct this but I’m unsure on what to do here. For _rolling , I added a check to exclude any -ve values while computing the average using _numIncrease so that should fix that bit"

clarification from me: the _numIncrease threshold was arbitrarily set; values can be negative if there are corrections to daily totals the next day. ideally, this threshold would be inversely proportional to the administrative level, so counties should see much smaller corrections compared to countries. large negative numIncreases should throw a warning rather than an error-- just a cue to examine the data to make sure there wasn't something weird going on with the geographic assignment, for instance.

[gkarthik]

Potentially relevant issue outbreak-info/outbreak.info#132.