Consider the situation of:
- The user submits a solution in a compiled language
- The user's solution compiles without error
- The first test passes, but the next test "produces an error" (for
  example, it raises an exception).

In this case, the correct top-level status is `fail`, with one
test-level status of `pass`, and one test-level status of `error`. But
the previous wording could imply that a top-level status of `fail`
is only correct if at least one test has the status `fail`.

The previous wording could imply that a per-test `message` property
can only be present when the test `status` is `fail`.

To me, the previous wording implied that a top-level `status` of `error`
is valid when at least one test "errors". But actually, if at least one
test was "executed", the top-level status should be `pass` or `fail`.

Co-authored-by: Erik Schierboom <erik_schierboom@hotmail.com>

Co-authored-by: Erik Schierboom <erik_schierboom@hotmail.com>

Co-authored-by: ee7 <45465154+ee7@users.noreply.github.com>