Bump default tuner_threads to 8

The previous default of 1 makes it too easy to run the tuner sequentially
and claim that TC is slow. Bumping to 8 CPU threads by default.
This is arguably unreasonable but if one cannot afford 8 threads for
autotuning TCs, one should not be in the business of running the tuner
in the first place.

Author: nicolasvasilache

tc/core/flags.cc

@@ -80,7 +80,7 @@ DEFINE_uint32(
     tuner_gen_number_elites,
     10,
     "The number of best candidates that are preserved intact between generations");
-DEFINE_uint32(tuner_threads, 1, "Number of CPU threads to use when autotuning");
+DEFINE_uint32(tuner_threads, 8, "Number of CPU threads to use when autotuning");
 DEFINE_string(
     tuner_devices,
     "0",