fix(recovery): recover from poisoned mutex in CircuitBreaker

[ashnaaseth2325-oss]

SUMMARY

This PR prevents runtime panics in the CircuitBreaker by safely handling poisoned mutex locks instead of unwrapping them. It updates lock acquisition logic across crates/mofa-foundation/src/recovery.rs, ensuring the circuit breaker continues functioning even after a thread panic.

---

FIX

• Root cause:
  std::sync::Mutex::lock().unwrap() panics when the mutex is poisoned, causing cascading failures in all subsequent calls.

• Technical approach:
  Replace .unwrap() with .unwrap_or_else(|e| e.into_inner()) to recover the guard from a poisoned mutex and continue execution safely.

Before (problematic pattern):

let mut guard = self.state.lock().unwrap();

After (safe handling):

let mut guard = self.state.lock().unwrap_or_else(|e| e.into_inner());

• Applied consistently across:

  • state()
  • call()
  • record_success()
  • record_failure()
  • reset()

---

VERIFICATION

1. Create a CircuitBreaker instance.
2. Simulate a panic while holding the mutex (e.g., inject a panic inside a guarded section like record_failure()).
3. After the panic, invoke methods like state() or call().

[ashnaaseth2325-oss]

Hello @LuigiGonnella @mugiwaraluffy56 ,
This PR fixes panic on poisoned mutex in CircuitBreaker by safely recovering the lock. It prevents cascading failures and keeps the circuit breaker working reliably.
Thanks!