Without requiring 'date', we get an exception in the retry middleware
when it tries to use DateTime for handling the Retry-After header.

Description

This simple script can be used to reproduce the bug:

#!/usr/bin/env ruby

require 'faraday'

conn = Faraday.new('https://amazon.com') do |f|
  f.request :retry, {
    retry_statuses: [301],
    max: 1,
    interval: 1.0,
    backoff_factor: 2
  }
end

conn.get('gp/css/order-history')

The fix is to simply add a require for Ruby's date library so we don't get an exception during retries.

I'm not certain this warrants a spec change (or if that'd even be possible?) but please let me know!