Strip raw output when unzipping benchmarks

[Chris Warburton]

We've decided to store gzipped benchmark results, rather than the raw,
uncompressed data, since it's so big. We have an 'unzipBenchmarks'
command to unzip this data into the right place for asv; but we still
can't generate HTML nicely from it, since the track_data benchmarks
don't play nicely with the graphing, etc.

We should alter the unzipBenchmarks script to replace any problematic
fields with dummy values. `null` would be a nice choice, since it's
small and conveys the meaning of "no data". If that doesn't work, we
might try removing the offending fields; although that might cause
problems if asv expects to find them due to e.g. `benchmarks.json`. If
so, a begrudging work around would be to stick zeroes in the fields...