Deal with overlapping timezones

[ateucher]

There are some regions in the world where a single point can land in
two different overlapping timezones. The "accurate" method includes these,
however the method used in the "fast" does not include overlapping timezones
(at least for now).

The implication is that the accurate method can return a result longer than the input, and since the return value is a character vector, there is no way to align the results to the input. Possibly make it a named vector with names like c(1, 2, 3.1, 3.2)...

[ateucher]

Could return a list where some elements are a single timezone, and where they are duplicated the list element is a length-2 vector of timezones. Or alternatively it always returns a data frame instead of a character vector, and throw a warning if the returned data frame is longer than the input