Rudimentary optimization with APM/QA.

Added script 'apm_quality_assessment_optimize' for finding parameters
that minimize a custom function of the scores generated by APM-QA. The
script reuses the existing functionality for filtering the data on
configs/scores/outputs.

To archieve that, some modularization has been done: the part from
apm_quality_assessment_export that reads in data into a
pandas.DataFrame has been moved into quality_assessment.collect_data.

TESTED = though extensive manual tests. Unit tests for the user
scripts and 'collect_data' are missing, because we don't have a test
framework for loading/exporting fake data.

BUG=webrtc:7218

Change-Id: I5521b952970243da05fc4db1b9feef87a2e5ccad
Reviewed-on: https://chromium-review.googlesource.com/643292
Commit-Queue: Alex Loiko <aleloi@webrtc.org>
Reviewed-by: Alessio Bazzica <alessiob@webrtc.org>
Cr-Commit-Position: refs/heads/master@{#19780}