explicitly use Dict.size rather than just size in equal?

Fixes bug where `true` was incorrectly returned for `Dict.equal?` when
the 1st argument was a subset of the 2nd argument.

e.g:
```elixir
iex> h1 = HashDict.new(foo: 1, bar: 2)
iex> h2 = HashDict.new(foo: 1, bar: 2, baz: 3)
iex> Dict.equal?(h1, h2)
true # OOPS
iex> Dict.equal?(h2, h1)
false # works as expected other way round
```

The code to check equality was checking size (incorrectly), and then
checking to see if every element in the first dict was present in the
2nd.

However, `size` was being used instead of `Dict.size`. This ended up
calling `Kernel.size` instead - and would always return `true`
regardless of whever the actual size of the `Dict` arguments were the
same or not.

Author: madlep

lib/elixir/lib/dict/behaviour.ex

@@ -99,7 +99,7 @@ defmodule Dict.Behaviour do
       defoverridable values: 1
 
       def equal?(dict1, dict2) do
-        case size(dict1) == size(dict2) do
+        case Dict.size(dict1) == Dict.size(dict2) do
           false -> false
           true -> 
             try do
@@ -122,4 +122,4 @@ defmodule Dict.Behaviour do
       defoverridable merge: 2, merge: 3
     end
   end
-end
+end

lib/elixir/test/elixir/dict_test.exs

@@ -214,6 +214,10 @@ defmodule DictTest.Common do
 
         dict2 = Dict.put(dict2, :a, 3)
         refute Dict.equal?(dict1, dict2)
+
+        dict3 = HashDict.new(a: 2, b: 3, f: 5, c: 123, z: 666)
+        refute Dict.equal?(dict1, dict3)
+        refute Dict.equal?(dict3, dict1)
       end
 
       test "unsupported dict" do