benchmarks/lockhammer/src/measure.c - measurement loop

The measurement functionality is moved to here from lockhammer.c.

The lock algorithm test header is #include inlined as before.  For each
test, a separate measurement.test.o file is made and linked against the
rest of the test harness object files to create an lh_test executable.

- when built to use the outline atomics library (USE_BUILTIN=1), the
  runtime flag --disable-outline-atomics-lse overrides the autodetection
  of LSE to run the exclusives routines instead

- blackhole and test thread synchronization routines are moved here

- work-based measurement (each thread completes N lock acquires) vs.
  time-based measurement (each thread runs for T hardware counter ticks)

- improve thread cleanup in case of crash or test timeout

- debug diagnostic to show per-thread workload progress against the
  hardware counter to show imbalance of workload execution between
  threads when work-based instead of time-based

Change-Id: Icfff17be52fa229e90ca23fc834de3fb187a88f7

Author: jty2