patch google symptom to account for dip

[aysim319]

between mid-May to mid-Aug 2024 there was a dip of data for google symptom; the dip was partially fixed and need to reflect the changes in covidcast

[nmdefries]

This is not worth patching.

The current/restored data has the same amount and placement of missingness as our current API data. (I initially thought that there were some values available in the new data that were missing in the API, but that was due to a date filter bug on my side (< vs <=).)

Some of the values of our composite signals (s01, etc) are a little different in the API vs using the new data, but the largest difference I saw at any geo level was O(1e-8), which is already outside of our 7-decimal digit rounding and insignificant for practical purposes.