Optimize Ruby 1.8 implementation of BSON::OrderedHash.

The set in there is unnecessary. Use the hash itself as "set". This makes
BSON::OrderedHash 148% faster on Ruby 1.8.

lib/bson/ordered_hash.rb

@@ -20,7 +20,6 @@
 #
 # Under Ruby 1.9 and greater, this class has no added methods because Ruby's
 # Hash already keeps its keys ordered by order of insertion.
-require 'set'
 
 module BSON
   class OrderedHash < Hash
@@ -60,7 +59,6 @@ module BSON
       def initialize(*a, &b)
         super
         @ordered_keys = []
-        @ordered_set = Set.new
       end
 
       def keys
@@ -69,10 +67,8 @@ module BSON
 
       def []=(key, value)
         @ordered_keys ||= []
-        @ordered_set ||= Set.new
-        unless @ordered_set.member?(key)
+        unless has_key?(key)
           @ordered_keys << key
-          @ordered_set.add(key)
         end
         super(key, value)
       end
@@ -103,7 +99,6 @@ module BSON
         @ordered_keys ||= []
         @ordered_keys += other.keys # unordered if not an BSON::OrderedHash
         @ordered_keys.uniq!
-        @ordered_set = Set.new(@ordered_keys)
         super(other)
       end
 
@@ -117,7 +112,6 @@ module BSON
 
       def delete(key, &block)
         @ordered_keys.delete(key) if @ordered_keys
-        @ordered_set.delete(key) if @ordered_set
         super
       end
 
@@ -137,7 +131,6 @@ module BSON
 
       def clear
         super
-        @ordered_set.clear
         @ordered_keys = []
       end