Can we get per-crate coverage data in a reasonable manner?

Previously we've been testing coverage via tarpaulin, which gives us per-crate coverage summaries. We're moving to grcov, which would rather give per-directory coverage summaries.

Is there a way we can extract per-crate coverage in a reasonable way? We've been aiming for "per-crate coverage above X%" as a benchmark in the past. Alternatively, we can just stick with per-directory coverage and define a new benchmark based on that.