RPC tests should support failure tests

[hanabi1224]

Issue summary

          I'm okay with the condition that it's explicit for this particular test (like `should_fail`). I wouldn't like the behaviour like before, where all matching internal server errors were considered a success.

Originally posted by @LesnyRumcajs in https://github.com/ChainSafe/forest/pull/4238#discussion_r1571999743

Other information and links