Incorrect handling for overlapping holiday dates

The generated_holidays table in R contains a non-existing holiday, Bevrijdingsdag, Hemelvaart, that is actually a concatenation of two holidays (Bevrijdingsdag and Hemelvaart) that happen to coincide in that year, but in general do not fall on the same date.
This has prophet fit a coefficient for this joint holiday instead of the separate holidays. A tidyr::separate_rows(holiday, sep = ", ") fixes this, but it should maybe be fixed in the source data instead.

> prophet::generated_holidays %>% filter(country == "NL", ds == "2005-05-05")

          ds                    holiday country year
1 2005-05-05 Bevrijdingsdag, Hemelvaart      NL 2005

[raffg]

Quick update, the Idaho MLK holiday has been updated in the beta branch of holidays: https://github.com/dr-prodigy/python-holidays/pull/354. I searched all countries and I believe this is the last remaining comma in a holiday name. Once this beta branch is merged with master, Prophet should not have any more issues with overlapping holiday dates.

[bletham]

Thanks for fixing this @raffg! I updated the R generated file too, so this should all be fixed (pending the upstream fix).