Zenguard

Introduce tiered API endpoints.

Fix the bugs in the
* OpenAI Integration
* Unit tests

New functionality added:

- Async Prompt Injection detection support.
- Prompt Injection detection reporting support.
- Added unit tests

Reworked:

- zenguard's detect methods now raise errors.

* If detector is single use a dedicated API
* Provide latency numbers
* Remove prints

Added running detectors in parallel.

Changed

- API change `message` is changed to `messages` and accepts the list of strings.