Make PyFluent Architecture Agentic-Ready

## I. Agentic AI Platform

### 1. Make PyFluent Architecture Agentic-Ready
We’ll need solid scaffolding to make Fluent a safe and robust substrate for **agentic workflows**.
Key enablers include:
- **Error transparency:** Extend the exception hierarchy to support fine-grained error handling.
- **Resilience:** Isolate errors on te client-side by enforcing pre-requisites (#4684)
- **Undo/redo:** Explore lightweight transaction support via `get_state()` snapshots around `set_state()`.
- **Constraints:** Expose enums and variable descriptors programmatically (ongoing). Investigate richer runtime constraints (valid ranges, dependencies).
- **Observability:** Enable tracing of *natural-language → PyFluent calls → Fluent actions* for explainability and debugging.
- **Sandboxing:** Investigate a future feature that would let users try out changes in a protected ‘test’ environment before anything becomes real. The system would check these trial changes for problems or inconsistencies, catching potential mistakes early and reducing the risk of breaking their actual setup.

No files or tests are named. Start by breaking the proposal into its listed areas: exception handling, prerequisites in #4684, get_state()/set_state() snapshots, descriptors and constraints, tracing, and sandboxing. Done requires an agreed architecture and implementation scope, but the issue does not identify a runnable validation path.

Written by the indexing model from the issue text.