Improve Coverage Measurement in FuzzBench

[jonathanmetzman]

We're planning on doing a lot of changes to improve coverage measurement in FuzzBench.
This will the tracking bug for those efforts.
@Shadoom7 can you mention this bugs in your pull requests so they can be tracked better.

[jonathanmetzman]

The first step in these efforts is integrating clang coverage with FuzzBench.
There are some questions about whether it is fast enough for this purpose.
@Shadoom7 will try replacing SanitizerCoverage with clang coverage on a branch, and then do experiments to compare the speed of measurement using clang coverage vs sanitizercoverage. Ideally we would not support both systems, but if clang coverage is too slow. We will only use it at the end of an experiment.

[jonathanmetzman]

Related to #77

[Shadoom7]

#564
This PR replaces the sanitizer coverage with clang coverage

[Shadoom7]

#565
This PR will collect and store the covered region data in cloud storage

[Shadoom7]

#595
This is the rebased version of #565.

[Shadoom7]

#602
Some fix to ensure the coverage info is always dumped.

[Shadoom7]

#618
Generate the clang coverage reports.

[Shadoom7]

#622
A fix to reduce the memory use during measuring.

[Shadoom7]

#657
Implemented two matrix and plots to measure and display the unique coverage of each pair of fuzzer and benchmark.

[Shadoom7]

#662
Add one column in the Experiment table in database to better support merging between experiments.

[Shadoom7]

#680
Fix the broken coverage reports.

[Shadoom7]

#685
Some fix and implemented one more summary ranking plot of unique coverage.