[red-knot] Improve property test performance by cloning db instead of holding MutexGuard (#16823)

## Summary

This PR brings an optimization.

- `get_cached_db` no longer returns a `MutexGuard`; instead, it returns
a cloned database.

### `get_cached_db`

Previously, the `MutexGuard` was held inside the property test function
(defined in the macro), which prevented multiple property tests from
running in parallel. More specifically, the program could only test one
random test case at a time, which likely caused a significant
bottleneck.

On my local machine, running:

```
QUICKCHECK_TESTS=100000 cargo test --release -p red_knot_python_semantic -- --ignored stable
```

showed about **a 75% speedup** (from \~60s to \~15s).

crates/red_knot_python_semantic/src/types/property_tests.rs

@@ -24,7 +24,7 @@
 //!   --ignored types::property_tests::stable; do :; done
 //! ```
 
-use std::sync::{Arc, Mutex, MutexGuard, OnceLock};
+use std::sync::{Arc, Mutex, OnceLock};
 
 use crate::db::tests::{setup_db, TestDb};
 use crate::symbol::{builtins_symbol, known_module_symbol};
@@ -327,9 +327,9 @@ impl Arbitrary for Ty {
 
 static CACHED_DB: OnceLock<Arc<Mutex<TestDb>>> = OnceLock::new();
 
-fn get_cached_db() -> MutexGuard<'static, TestDb> {
+fn get_cached_db() -> TestDb {
     let db = CACHED_DB.get_or_init(|| Arc::new(Mutex::new(setup_db())));
-    db.lock().unwrap()
+    db.lock().unwrap().clone()
 }
 
 /// A macro to define a property test for types.
@@ -348,8 +348,7 @@ macro_rules! type_property_test {
         #[quickcheck_macros::quickcheck]
         #[ignore]
         fn $test_name($($types: super::Ty),+) -> bool {
-            let db_cached = super::get_cached_db();
-            let $db = &*db_cached;
+            let $db = &super::get_cached_db();
             $(let $types = $types.into_type($db);)+
 
             $property