Re-commit "Make dwarf parsing multi-threaded"

Re-commit the change after fixing a lot of race condition in LLDB
exposed by this change

Loading the debug info from a large application is the slowest task
LLDB do. This CL makes most of the dwarf parsing code multi-threaded.

As a result the speed of "attach; backtrace; exit;" when the inferior
is an LLDB with full debug info increased by a factor of 2.

Differential revision: http://reviews.llvm.org/D13662

llvm-svn: 251106