Curl::Multi.http behaves incorrectly when using the :max_connects option

[sriedel]

It seems that it re-uses the urls from the restart of the list:

2.2.3 :001 > require 'curb'
 => true 
2.2.3 :002 > results = []
 => [] 
2.2.3 :003 > Curl::Multi.http( [ { :url => "http://www.google.de", :method => :get }, { :url => "http://www.cnn.com", :method => :get }, { :url => "http://ft.com", :method => :get } ], :max_connects => 1 ) { |*args| results << args }
 => nil 
2.2.3 :004 > results
 => [[#<Curl::Easy http://www.google.de>, nil, :get], [#<Curl::Easy http://www.google.de>, nil, :get], [#<Curl::Easy http://www.google.de>, nil, :get]] 
2.2.3 :007 > results.clear
 => [] 
2.2.3 :008 > Curl::Multi.http( [ { :url => "http://www.google.de", :method => :get }, { :url => "http://www.cnn.com", :method => :get }, { :url => "http://ft.com", :method => :get } ], :max_connects => 2  ) { |*args| results << args }
 => nil 
2.2.3 :009 > results
 => [[#<Curl::Easy http://www.google.de>, nil, :get], [#<Curl::Easy http://www.google.de>, nil, :get], [#<Curl::Easy http://ft.com>, nil, :get]] 
2.2.3 :010 > results.clear
 => [] 
2.2.3 :011 > Curl::Multi.http( [ { :url => "http://www.google.de", :method => :get }, { :url => "http://www.cnn.com", :method => :get }, { :url => "http://ft.com", :method => :get } ], :max_connects => 3  ) { |*args| results << args }
 => true 
2.2.3 :012 > results
 => [[#<Curl::Easy http://www.cnn.com>, nil, :get], [#<Curl::Easy http://www.google.de>, nil, :get], [#<Curl::Easy http://ft.com>, nil, :get]]