Unexpected supports_protocol behavior

We use `supports_protocol` to check if objects being handled can be adapted to a specific protocol and use the information appropriately.

Trying to adapt an object to a specific protocol can result in three behaviors

- the object is adaptatable, in which case `supports_protocol` returns `True`
- the object is unadaptable because no appropriate adapters are registered with the adaptation machinery, in which case `supports_protocol` returns `False`
- the object is unadaptable because an `AdaptationError` occurs during adaptation.
- the object is unadaptable because other `Exception`s occur during adaptation.

At the moment, the `AdaptationError`s are not handled by `supports_protocol`, which we believe they should. The adapters registered could be explicitly raising `AdaptationError`s - which should be caught by `supports_protocol`, which would then return `False`.

https://github.com/enthought/traits/blob/5dddc966b8f87e5550e83c98e14b4a251b33b561/traits/adaptation/adaptation_manager.py#L183-L191

Start in traits/adaptation/adaptation_manager.py at lines 183-191, linked from the issue, and trace how supports_protocol handles adaptation failures. Verify the behavior for adapters that raise AdaptationError, and consider the existing adaptation-related tests when checking the result. Done means those cases return False without changing the handling of other exceptions.

Written by the indexing model from the issue text.