New Design MutliMethod Hot Method Recompilation

As part of my GSOC I proposed 3 parallel compilation strategies single-method, multi-method, multi-phase.

Single-method is done however it is only relevant for large methods and would easily be overshadowed by multi-phase in a match up. However, multi-method is relevant for all cases and even our worst case benchmarks show speedup.

Here are the results numerically

2 Threads
Best Case LU: 816.5870
AVG Case LU: 795.67

Original
Best Case LU : 805.98
AVG Case LU: 793.05

2 Threads
for I in {1..90}; do ./rvm -X:aos:num_threads=2 -cp scimark2lib.jar jnt.scimark2.commandline; done