Improve optimization of operations on indexables by using more inferred type information and by dis…

Improve optimization of operations on indexables by using more inferred type information and by disabling inlining of [], []= and length (still optimized but now in the folder).

R=ngeoffray@google.com
BUG=

Committed: https://code.google.com/p/dart/source/detail?r=21946

[ngeoffray, 2013-04-24 12:55:01]

LGTM

[kasperl, 2013-04-24 13:03:48]

Committed patchset #3 manually as r21946 (presubmit successful).

[vsm, 2013-04-25 00:07:08]

https://codereview.chromium.org/14052009/diff/5001/tests/html/html.status
File tests/html/html.status (left):

https://codereview.chromium.org/14052009/diff/5001/tests/html/html.status#old...
tests/html/html.status:23: typed_arrays_range_checks_test: Fail # TODO(vsm):
Implement range checks.
Kasper - I was just looking into adding explicit bounds checks here.  Are they
no longer needed?  Looks like you're special casing the typed lists now.

[kasperl, 2013-04-25 05:24:31]

https://codereview.chromium.org/14052009/diff/5001/tests/html/html.status
File tests/html/html.status (left):

https://codereview.chromium.org/14052009/diff/5001/tests/html/html.status#old...
tests/html/html.status:23: typed_arrays_range_checks_test: Fail # TODO(vsm):
Implement range checks.
On 2013/04/25 00:07:08, vsm wrote:
> Kasper - I was just looking into adding explicit bounds checks here.  Are they
> no longer needed?  Looks like you're special casing the typed lists now.

There's still more to do. What I added in this CL gives us index bounds checks
for typed data lists whenever we optimize the access to them -- if we have too
little type information available to us to optimize the access, we will still
call the [] and []= operators on the typed lists.

It's a little bit weird, but we now have index checks in the optimized case -
and no such checks in the unoptimized case.