Fix imports from dplyr and lubridate

[capnrefsmmat]

Bizarrely, even though R CMD check passed perfectly, the uses of dplyr and lubridate introduced in #275 didn't actually work in the installed package:

> library(covidcast)
> symptoms <- covidcast_signal("fb-survey", "smoothed_hh_cmnty_cli", geo_type="state")
Fetched day 2020-04-15 to 2020-06-23: 1, success, num_entries = 3601
Error in summarize(., geo_value = list(geo_value)) : 
  could not find function "summarize"
> # and after fixing that:
> foo = covidcast_signal("fb-survey", "smoothed_hh_cmnty_cli", start_day="2020-11-20", end_day="2020-11-24", geo_type="state")
Fetched day 2020-11-20 to 2020-11-24: 1, success, num_entries = 153
Error in ymd(missing_time_values) : could not find function "ymd"

This is because there was no @importFrom lubridate ymd or equivalent, or the same for dplyr, so these functions are not in the package NAMESPACE.

I've fixed dplyr here by using the dplyr:: prefix, and simply replaced the ymd() calls with a short function, since we were already using as.Date() for the same purpose in several places.

I'm not sure why the tests worked -- this is exactly the kind of problem testing should have detected. Maybe because some of the test files use dplyr, it was in the namespace, and none of the tests exercise the code that called ymd()? Dunno.

[JedGrabman]

Is there an open issue for investigating why R CMD check succeeded and if not, can you create one? We should try to prevent this sort of problem from being able to happen again.