Adopt prefixdate library for parsing dates

[pudo]

This would adopt the prefixdate library, which makes some of the logic of the existing date parsing more explicit. When applied, this should:

a) Speed up date processing, both for format-string based parsing and for normal validation.
b) Truncate prefixes for minutes and seconds from times when they are 00.
c) Fix an apparent issue with time zone conversion in to_datetime.

[Rosencrantz]

Alright, just so I understand what you're proposing here. You're looking to remove the big, ugly Regex with a library that does the same thing, and this is why most of this PR is removing code?

[pudo]

Haha, I guess the point is that the library uses one fat regex to do a lot of the stuff that's happening in individual steps here (like that weird CUT_ZEROES thing)...

One thing that's a bit "complicated" is that prefixdate disregards unknown suffixes on dates that it does not understand. That means 2001BANANA is a valid date, and will normalise to 2001. That's needed to pass the existing ftm tests, which contain some weird examples. Maybe we could at some point adapt and get more strict.

[pudo]

OK, the time issue is fixed.