Added PTH optimization to not process entire blocks of tokens that appear in skipped preprocessor blocks.  This improves PTH speed by 6%.  The code for this optimization itself is not very optimized, and will get cleaned up.

llvm-svn: 60956