simpleperf: speed up inferno.

1. Add --min_callchain_percentage option. This threshold is used to trim
unimportant callchains. This avoids too much time generating the report
when processing millions of samples.

2. Improve some bottleneck places shown in cProfile.

Bug: 66914187
Test: run test.py.
Change-Id: I64347ee3c32813166736ebde699cbed8b129183d
3 files changed