Add data scraper for Contra Costa County

[Mr0grog]

There are quite a few complexities and caveats with this one:

• Positive test counts are estimated based on rate (since the county only publishes the rate), similar to Santa Clara.

• Daily test counts, total test counts, and test positivity rates are each separate charts that cover different date ranges! Putting them together as a single series is a bit messy. Worse, daily test counts and totals don't completely agree (the total tests series starts earlier, and roughly the first two weeks of daily tests have 0 for the value, while the difference between total tests on those days is often > 0). After March 18, 2020 they mostly agree, so daily total test counts seems like the better source. I've calculated daily tests based on the difference between days in total tests, rather than using the daily counts directly.

• Contra Costa splits out hispanic ethnicity from the other racial categories, unlike all the other counties, which provide them as a combined cross section. I'm not sure there's a good way to make our own combined version, so I added an new grouping for ethnicity, and left the other races in race_eth as usual. It's not great for comparing to other counties. (Maybe we could assume "hispanic" is a chunk out of "white" and calculate the cross section that way, but that seems pretty iffy. I'm open to feedback or suggestions.)

• Case race and case ethnicity are published as percentages, unlike the other case demographics and the corresponding death demographics, so I've calculated them based on the totals. Kind of a weird exception (especially since the same thing isn't done for deaths).

Fixes #21.

[Mr0grog]

Rebased on master for a cleaner merge.