[merge] Re: [bug] if first running test miss dependency then selftest fails horribly

[Martin Pool]

On consideration, although this does fix the problem Alexander
reported, I think it's papering over the way the _benchmarkTime field
of ExtendedTestResult is used.

We have just one ExtendedTestResult per run, I think, and every test
is passed into it.  So it's a bit bogus to have fields on it that
apply just to the most recent test we looked at, unless they're
clearly marked as such.

-- 
Martin