Please evaluate draft fix for RT-33390

[Jim Graham]

A number of recent bugs have stemmed from a mistaken assumption in the 
D3D code that calls to update the render target would always cause a new 
render target to be installed which would clear the clip on D3D. 
Recently, that assumption started failing and sometimes the render 
target does not change and so the clip is not cleared.

The following patch was developed to try to fix these bugs, but I could 
not reproduce all of the original problems to know if this fix even 
targets all of these bugs.  I need the submitters of the various bugs to 
evaluate the patch to see if it fixes their problems:

webrev: http://cr.openjdk.java.net/~flar/RT-33390/webrev.00/

Should fix:

https://javafx-jira.kenai.com/browse/RT-34229
This was the original bug I was trying to fix and I've verified the 
patch fixes this bug.

https://javafx-jira.kenai.com/browse/RT-34159
It looks like Steve has already verified that it fixes this bug as well.

https://javafx-jira.kenai.com/browse/RT-33390
This is the bug that I could not reproduce to verify the fix.  I'd 
appreciate someone who saw the problem to verify the patch (attached to 
this Jira issue as a patch file).

Finally, some feedback on the ugly way that I pass through a boolean 
array to communicate the clearing of the clip back to the Java level. 
Is there another way to achieve this result in the D3D module?

			...jim