Perf summaries ignore max-rss #988
Description
Max-rss has historically been noisy for rustc, but it's been quite accurate for rustdoc. rust-lang/rust#88522 (comment) says "Summary: This benchmark run did not return any relevant changes.", but the max-rss measurement shows a consistent 1% improvement on -doc runs. It would be great for the summary to reflect that.
(This could be a duplicate of #980, but I suspect you went with just instruction count for the summaries while implementing the prototype.)