Add support for decoding timestamp from existing ULID

[cocoahero]

This is another attempt at @joshbeckman's original PR (#31) which adds support for decoding the timestamp portion of an existing ULID value.

The main difference is this version fixes the flakey spec by using to_r instead of to_f when comparing the two timestamp values. I used .to_r for the same reasons as already noted here: https://github.com/rafaelsales/ulid/blob/master/lib/ulid/generator.rb#L59-L69

[dmke]

I am in need of this being a part of the gem :)

[dmke]

ULIDs are case-insensitive. This currently fails:

it 'decodes in a case-insensitve manner' do
  ulid = ULID.generate.downcase
  assert ULID.decode_time(ulid).is_a?(Time)
end

For completeness, I'd like to see a few known input/output pairs (but please don't feel obliged to include them):

it 'correctly parses known test vectors' do
  {
    "0000000000Y2GBSG3FEFT635J1" => Time.at(0),
    "013XRZP292318JWRM98F0YAPV9" => Time.at(1234567891234/1000r),
    "ZZZZZZZZZZ-this-is-ignored" => Time.at(1125899906842623/1000r),
  }.each do |vector, expected_time|
    assert_equal expected_time, ULID.decode_time(vector)
  end
end