Solve all isodate issues

[bolsote]

Given one of the main motivations of this library was to improve the approach taken by isodate, we should trawl their issue tracker, and make sure we actually solve the most important problems found there.

[bolsote]

A quick overview of the isodate issue tracker:

I do believe we would not be affected by these (although we will double check):

• Parse time interval gweis/isodate#48 would be sort of related to Implement repeating intervals #6. We will be ignoring it for now.
• I have some doubts about Allow lower case letters. gweis/isodate#52. I'll have to check the standard wording myself and check the state of the art.
• I don't think we should really be affected by anything like parse_time() is not threadsafe gweis/isodate#54, but a quick check would not go astray.
• Part of the reason for the original design was to avoid parse_datetime() is slow gweis/isodate#9. Our benchmarks show we are indeed faster.
• Our type system looks the way it is because of isodate.parse_datetime incorrectly handles Years, Months and Weeks gweis/isodate#42. Our internal representation should also shield us from Fractional duration gweis/isodate#2. Wrong datetime return gweis/isodate#21 refers to datetime parsing, which we do not do here.
• Our hand-rolled parser handles errors in a very different way from isodate, so these should not apply, save for programmer error:
  • isodate.duration_isoformat does not raise Exception for wrong input type gweis/isodate#56
  • parse_date do not raise Exception on incorrect string (GPS format) gweis/isodate#63
  • Duration includes weeks in combination with other units of time gweis/isodate#27
  • parse_datetime can raise ValueError gweis/isodate#13
  • Parser is too lax gweis/isodate#15
• We do support negative durations, so Negative duration values gweis/isodate#41 should not apply.

I'm drawing a blank on gweis/isodate#53. Writing a test should clarify what that's about.

[bolsote]

The way I read it what's requested in gweis/isodate#52 would be incompatible with both the wording of ISO/DIS 8601-1, and with the EBNF grammar provided in Appendix A of said document.