`peer_return()` and un-knit Rmd files for review

[mine-cetinkaya-rundel]

Reviewers are asked to put their reviews in a *-review.Rmd file. Then, in the peer_return() step this is the file that is linked in the issue on the author's repo for them to see their feedback. We probably want to move the *-review.md file instead of the Rmd and link to that.

This might be an easy fix, except...

Some reviewers do not knit their file, resulting in no md output in the review repos. We could solve this by running an action on the review repos but then we need to have something in place for what if the file doesn't knit -- fall back to Rmd?