This patch rewrites common_invokeOld for the x86 fast interpreter. The implementation...

This patch rewrites common_invokeOld for the x86 fast interpreter. The implementation is similar to what is done for the armv5 code.
Testing shows the patch provides a performance boost to benchmarks such as Caffeine Mark. When testing the simulator build on an Intel(R) Core(TM)2 Quad  CPU   Q9550  @ 2.83GHz, the string score for Caffeine Mark improved by 6%, the Method score improved by 12% and the overall score improved by 3%.

This patch is one of our incremental efforts to merge some Intel's x86 fast interpreter features in http://review.source.android.com/Gerrit#change,6041 to the mterp/x86 directory.