Catch failures & show as did not complete?

I’m trying to benchmark my library against an existing one. It can’t complete the bench since it doesn’t implement the output correctly. If I add `failwith` the whole benchmark suite fails, but I think it should show some a “did not complete message”.

No source file or test is named in the issue. Start by locating the benchmark execution and failure-handling entry points, then reproduce a benchmark that raises failwith; done means an incomplete benchmark is reported as “did not complete” without failing the whole suite.

Written by the indexing model from the issue text.