document how to get a cachegrind diff from a perf run

[jyn514]

it shows a useful command in the 'detailed-query' view: https://perf.rust-lang.org/detailed-query.html?commit=4b87ed9d0f18166a26aadb5289bac004e058b90f&base_commit=2034b6d23ca18311a23273f2f08c381a04fa05d9&benchmark=serde_derive-1.0.136-check&scenario=incr-unchanged and there's instructions on zulip for how to build from source to get debuginfo: https://rust-lang.zulipchat.com/#narrow/stream/247081-t-compiler.2Fperformance/topic/annotations.20for.20cachegrind.20diff.3F

[Mattias-Petersson]

@rustbot claim