The prime focus of this PR is to catch errors during iterations properly and write the results just like we do for non-errored benchmarks. Along the way this addresses some issues this uncovered, and makes a few judgment calls about what goes where, so comments are welcome!
Ok, I think this is polished and ready to go, but holding off on merging till the accompanying changes to labs/benchmarks are finished in case I need any other tweaks here.
Necessary for addressing https://github.com/voltrondata-labs/benchmarks/issues/134; a separate (coming) labs/benchmarks PR will be necessary to pick up all the data this is now sending properly.
The prime focus of this PR is to catch errors during iterations properly and write the results just like we do for non-errored benchmarks. Along the way this addresses some issues this uncovered, and makes a few judgment calls about what goes where, so comments are welcome!