ops[tracing] consider rewriting the vendored cham lib in pure Python

[dimaqq]

We've had discussions if it's worth it, given how prevalent pydantic is among charmers.

Why would we want to do this and what could be in scope:

• smaller potential 3rd party dep incompatibility surface
• faster build times for charms that are traced, but don't use (heavy? many?) relation charm libs
• much smaller codebase, if we remove the Provider
• much simpler codebase, if we streamline the Requirer and hardcode e.g. the protocol
• the charm lib can be architected much better
• avoids calling hook tools in __init__

To do:

• [ ] drop the pydantic dep
• [ ] update the docs