Currently, the overlapping results of individual benchmark tests are overwritten.

Example:
  tools/perf-compare.py Result,ss.json,kr.json Master,kr-m.json,ss-m.json -o compare.html
or
  tools/perf-compare.py Result,ss.json Result,kr.json Master,kr-m.json Master,ss-m.json -o compare.html

BUG=v8:6144
NOTRY=true

Change-Id: Ia340e8c01c46da17b1f4eee6c6bb8e5e9bb12c3e
Reviewed-on: https://chromium-review.googlesource.com/459537
Commit-Queue: Igor Sheludko <ishell@chromium.org>
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44090}

Inspired by bsuite/compare-baseline.py and tools/perf-to-html.py. The main difference
from the latter is that the new script does not require json files preprocessing and it
is able to compare multiple json files at once (just like bsuite/compare-baseline.py).

This CL also makes tools/run_perf.py relative path friendly.

BUG=

Change-Id: I7be796fa920f8360703806f32767b992eec0f9c2
Reviewed-on: https://chromium-review.googlesource.com/456286Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Commit-Queue: Igor Sheludko <ishell@chromium.org>
Cr-Commit-Position: refs/heads/master@{#43982}