In manual proof mode, interpret to point / step backward should interrupt computation if necessary

[JasonGross]

Currently, if I accidentally evaluate a looping computation, there doesn't seem to be a way to interrupt it without reloading the window / killing the process. This makes working in files with long proofs essentially infeasible.

[rtetley]

Thanks for reporting ! This is a duplicate of #373 and there is currently an open PR working on it (#1046)