Add [OverrideBuiltins] to HTMLCollection IDL interface

LayoutTests/fast/dom/collection-length-should-not-be-overridden-expected.txt

Index: LayoutTests/fast/dom/collection-length-should-not-be-overridden-expected.txt
diff --git a/LayoutTests/fast/dom/collection-length-should-not-be-overridden-expected.txt b/LayoutTests/fast/dom/collection-length-should-not-be-overridden-expected.txt
index 9cf751c28b8eed981e6093cf1bfb710b30c70ab3..e3f8328bda999b16f603da8ec375ece7ce633fbb 100644
--- a/LayoutTests/fast/dom/collection-length-should-not-be-overridden-expected.txt
+++ b/LayoutTests/fast/dom/collection-length-should-not-be-overridden-expected.txt
@@ -13,7 +13,7 @@ PASS form.elements[1].name is 'length'
 PASS select.options.length is 2
 PASS select.options[0].getAttribute('name') is 'length'
 PASS select.options[1] is select.options['some']
-PASS table.rows.length is 2
+FAIL table.rows.length should be 2 (of type number). Was [object HTMLTableRowElement] (of type object).

[arv (Not doing code reviews), 2014/02/05 20:53:05]

Speed over correctness makes me sad :'(

[Inactive, 2014/02/05 20:58:47]

I know... To be fair, we are returning HTMLCollections instead of NodeList for
getElementByClassName() / getElementByTagName*() now, which is good for
correctness (ditto for NodeList not having a named getter anymore). Sadly, the
regression caused by this is big enough that I don't think we want to live with
it. So, short term, it seems to be either this (Adding [OverrideBuiltins] to
HTMLCollection) or reverting my big patch to return HTMLCollections (which would
make me sad TBH).

I am open to other ideas though :)

 PASS table.rows[0] is table.rows['another']
 PASS table.rows[1].id is 'length'
 PASS successfullyParsed is true