Add masking shape classifier and guard tests for adversarial resistance

- Implemented tests for masking shape classifier resistance against threshold attacks, ensuring that blurring reduces accuracy and increases overlap between classes.
- Added tests for masking shape guard functionality, verifying that it maintains expected behavior under various conditions, including timeout paths and clean EOF scenarios.
- Introduced helper functions for calculating accuracy and handling timing samples to support the new tests.
- Ensured that the masking shape hardening configuration is properly utilized in tests to validate its effectiveness.

src/proxy/tests/client_masking_diagnostics_security_tests.rs

@@ -95,6 +95,7 @@ async fn capture_forwarded_len(body_sent: usize) -> usize {
     cfg.censorship.mask_unix_sock = None;
     cfg.censorship.mask_host = Some("127.0.0.1".to_string());
     cfg.censorship.mask_port = backend_addr.port();
+    cfg.censorship.mask_shape_hardening = false;
     cfg.timeouts.client_handshake = 1;
 
     let accept_task = tokio::spawn(async move {