USA source revison

[akarlinsky]

A user email alerted me to a small issue in US data: There's no 2015W1 and 2017W1 shows up twice.

This was due to combining of STMF data (which doesn't have 2015W1) with CDC data.

For the US, we have switched to using CDC data throughout. README has been updated accordingly:
United States (weekly)
2015--2017: Centers for Disease Control and Prevention

From 2017: Centers for Disease Control and Prevention: https://data.cdc.gov/api/views/xkkf-xrst/rows.csv
We use the 'predicted' (weighted) time series that accounts for underreporting in recent weeks. See https://www.cdc.gov/nchs/nvss/vsrr/covid19/excess_deaths.htm for more information.
We remove the last weeks (usually two) that are marked at https://gis.cdc.gov/grasp/fluview/mortality.html as being <90% complete.

[akarlinsky]

Commit 08774d7 fixed this.
I'm leaving this issue open for a few days for sake of public record.