faster es53 freeze

This change reduces runtime of a minimal test gadget
from 2800ms to 2100ms in Chrome Beta (16.0),
from 2300ms to 1800ms in Firefox Aurora (10.0),
from 2300ms to 2100ms in Firefox Beta (9.0),
from 1700ms to 1200ms in Safari 5.1.2

Basically, a much of the cost of es53+domado startup is freeze,
and much of the cost of freeze is iterating over object properties,
which includes inherited properties which have to be ignored.

This change tries to use a native Object.keys function to do
a fast initial filtering of the properties.

[MarkM, 2011-12-14 07:00:57]

http://codereview.appspot.com/5484074/diff/1/src/com/google/caja/es53.js
File src/com/google/caja/es53.js (right):

http://codereview.appspot.com/5484074/diff/1/src/com/google/caja/es53.js#newc...
src/com/google/caja/es53.js:226: // Due to IE<9 bugs, this isn't identical to
Object.keys, but we're
What are these IE<9 bugs?

[felix8a, 2011-12-14 07:25:35]

On Tue, Dec 13, 2011 at 23:00, <erights@gmail.com> wrote:

>
> http://codereview.appspot.com/**5484074/diff/1/src/com/google/**
>
caja/es53.js<http://codereview.appspot.com/5484074/diff/1/src/com/google/caja/es53.js>
> File src/com/google/caja/es53.js (right):
>
> http://codereview.appspot.com/**5484074/diff/1/src/com/google/**
>
caja/es53.js#newcode226<http://codereview.appspot.com/5484074/diff/1/src/com/google/caja/es53.js#newcode226>
> src/com/google/caja/es53.js:**226: // Due to IE<9 bugs, this isn't
>
> identical to Object.keys, but we're
> What are these IE<9 bugs?
>

a non-enumerable property makes descendent shadows non-enumerable.
like if I set o.toString, it doesn't show up in any enumeration of o.

so, that function I wrote is not identical to Object.keys because it's
potentially missing some keys that should be enumerable, but those
keys would be ignored anyway by our code, so it doesn't matter.

[felix8a, 2011-12-14 17:51:09]

new snapshot, clarified the IE<9 comment.

[felix8a, 2011-12-14 18:09:20]

new snapshot, a few more micro-optimizations for freeze

[felix8a, 2011-12-14 18:38:22]

On 2011/12/14 18:09:20, felix8a wrote:
> new snapshot, a few more micro-optimizations for freeze

now my minimal gadget test takes 1800ms to run in Chrome, vs 1600ms with a
neutered freeze.

[metaweta, 2011-12-14 18:52:29]

lgtm

[felix8a, 2011-12-14 23:07:02]

@r4683