Make speed_benchmark_torch report latency in us (#37953)
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/37953
Earlier it said us but reported ms.
Test Plan: buck run aibench:run_bench -- -b aibench/specifications/models/pytorch/fbnet/fbnet_mobile_inference.json --devices s9u --remote --framework pytorch --logger_level info --job_queue aibench_interactive --platform android/full_jit
Reviewed By: xcheng16
Differential Revision: D21349612
fbshipit-source-id: b97b6216eb0264123ff2c7852a0678b2008b0bf1
