Add short_metrics

[JackCaoG]

This pr added an api short_metrics_report which by defualt will emit a very short list of relevant metrics. User can also customize which metrics they want to see. IMO full metrics is way too detail in most time.

We will always print out the aten:: counters since they indicate fallback and are important to users.

example output:

Metric: CompileTime
  TotalSamples: 2
  Accumulator: 059ms522.421us
  ValueRate: 01s330ms003.908us / second
  Rate: 45.4528 / second
  Percentiles: 1%=026ms399.367us; 5%=026ms399.367us; 10%=026ms399.367us; 20%=026ms399.367us; 50%=032ms123.054us; 80%=032ms123.054us; 90%=032ms123.054us; 95%=032ms123.054us; 99%=032ms123.054us
Metric: ExecuteTime
  TotalSamples: 3
  Accumulator: 017ms966.371us
  ValueRate: 322ms039.563us / second
  Rate: 56.9432 / second
  Percentiles: 1%=001ms307.637us; 5%=001ms307.637us; 10%=001ms307.637us; 20%=001ms307.637us; 50%=008ms736.607us; 80%=008ms922.127us; 90%=008ms922.127us; 95%=008ms922.127us; 99%=008ms922.127us
Metric: TransferToServerTime
  TotalSamples: 1
  Accumulator: 002ms847.252us
  Percentiles: 1%=002ms847.252us; 5%=002ms847.252us; 10%=002ms847.252us; 20%=002ms847.252us; 50%=002ms847.252us; 80%=002ms847.252us; 90%=002ms847.252us; 95%=002ms847.252us; 99%=002ms847.252us
Metric: TransferFromServerTime
  TotalSamples: 3
  Accumulator: 013ms067.510us
  ValueRate: 279ms381.785us / second
  Rate: 64.1396 / second
  Percentiles: 1%=850.899us; 5%=850.899us; 10%=850.899us; 20%=850.899us; 50%=005ms494.984us; 80%=007ms721.627us; 90%=007ms721.627us; 95%=007ms721.627us; 99%=007ms721.627us
Counter: CachedCompile
  Value: 1
Counter: MarkStep
  Value: 2
Counter: aten::_local_scalar_dense
  Value: 1

[will-cromar]

Otherwise LGTM