AI observability using OpenInference and Phoenix

[hype-eater]

Adds basic OpenInference OTel tracing, using locally running Phoenix as the observability backend.

Note:
The code assumes you are using Phoenix as your backend (you could actually send the traces to any OTel-compliant backend). Also, the expectation is that Phoenix (or whatever backend you like) is already running (as in the README).

Alternatively:
For demos and workshops it might be simpler just to start Phoenix from within the app at start up, as in:

import phoenix as px
session = px.launch_app()  # for file-based persistence, pass use_temp_dir=False