forked from catchorg/Catch2
This stops tests failing falsely if the assertion passed, but the stringification itself failed as the assertion was sent to the reporter. I don't think that stringification should be fallible, but the overhead in compilation ended up being small enough (<0.5% on `SelfTest`) that it might be worth implementing, in case there is more users with weird `StringMaker`s than just MongoDB. Closes #2980