Speed up non-interceptor case of Object.getOwnPropertyNames

Speed up non-interceptor case of Object.getOwnPropertyNames

When there are interceptors on an object, it's possible to
end up with duplicate property names. But when all the names
are provided by v8, a collision is not possible, so we can
fast-path that case by not de-duping.

Also added better test coverage for interceptor API.

Committed: https://code.google.com/p/v8/source/detail?r=13725

[adamk, 2013-02-23 00:14:32]

Motivation here was that Rafael was seeing getOwnPropertyNames show up in a
profile and I wondered why it might be slow.

[Michael Starzinger, 2013-02-25 11:57:07]

LGTM, if it improves the test case that Rafael was looking into.

[adamk, 2013-02-25 17:23:37]

On 2013/02/25 11:57:07, Michael Starzinger wrote:
> LGTM, if it improves the test case that Rafael was looking into.

Yep, sped up his code by ~50% I think, making an iteration over
getOwnPropertyNames similar in speed to using a for-in loop (which obviously has
different semantics).

[adamk, 2013-02-25 18:58:53]

Committed patchset #2 manually as r13725 (presubmit successful).

[Vyacheslav Egorov (Google), 2013-02-26 00:12:31]

DBC (I known that it is not your code, but I noticed it while reading this CL)

Array.prototype.concat = function () { return ['ha', 'haha']; };
print(Object.getOwnPropertyNames({}).join(',')); // => ha,haha

[adamk, 2013-02-26 00:55:00]

On 2013/02/26 00:12:31, Vyacheslav Egorov (Google) wrote:
> DBC (I known that it is not your code, but I noticed it while reading this CL)
> 
> Array.prototype.concat = function () { return ['ha', 'haha']; };
> print(Object.getOwnPropertyNames({}).join(',')); // => ha,haha

Yep, noticed that as I was editing the code and thought I'd better fix that at
some point. Why not now? See https://codereview.chromium.org/12342003/