Allow expecting a parameter to be nil during return_to verification.

[pierre-pretorius]

There was a minor openid change on Google's side that caused all our applications using the ruby-openid gem to fail when authorizing users via Google openid. After some debugging I determined the return_to parameter in the response from Google changed from:

{"user[email]"=>[""], "user[password]"=>[""], "user[identity_url]"=>["https://www.google.com/accounts/o8/id"]}

to:

{"user[email]"=>[], "user[password]"=>[], "user[identity_url]"=>["https://www.google.com/accounts/o8/id"]}

This causes the verify_return_to_args method to reject the response even though the rt_key (expected result) is nil.

if msg_val.nil?
  raise ProtocolError, "Message missing return_to argument '#{rt_key}'"
elsif msg_val != rt_val
  raise ProtocolError, ("Parameter '#{rt_key}' value "\
                                      "#{msg_val.inspect} does not match "\
                                      "return_to's value #{rt_val.inspect}")
end

Please accept my pull request that addresses this problem. It doesn't reject the response if the rt_key (expected result) is also nil.

[dennisreimann]

Thanks for your contribution - I just pulled in those changes and pushed v2.4.0 to rubygems :)