Add key.success? and key.failure? predicates

In complex rules, it may be useful to know whether the key has any failure message already.

```ruby
rule(:foos).each do
key.success? # => true
key.failure? # => false
key.failure('wrong')
key.success? # => false
key.failure? # => true
end
```

Begin by tracing the key object exposed inside rule(:foos).each and how failure messages are recorded. Completion should cover the example's state transitions: success? is true before a failure, then false, while failure? changes from false to true after key.failure('wrong'). Add focused tests for these transitions.

Written by the indexing model from the issue text.