fix: replace tracing-android with android_logger (no sharded_slab SIGSEGV)

tracing_subscriber::registry() allocates a sharded_slab which causes
SIGSEGV on Android 16 MTE devices during nativeInit. catch_unwind
can't catch SIGSEGV (it's a signal, not a panic).

Replace with android_logger (lightweight, no large allocations) +
tracing-log bridge so tracing::info! macros still work via logcat.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

crates/wzp-android/Cargo.toml

@@ -17,7 +17,6 @@ wzp-crypto = { workspace = true }
 wzp-transport = { workspace = true }
 tokio = { workspace = true }
 tracing = { workspace = true }
-tracing-subscriber = { workspace = true, features = ["env-filter"] }
 bytes = { workspace = true }
 serde = { workspace = true }
 serde_json = "1"
@@ -28,7 +27,9 @@ libc = "0.2"
 jni = { version = "0.21", default-features = false }
 rand = { workspace = true }
 rustls = { version = "0.23", default-features = false, features = ["ring"] }
-tracing-android = "0.2"
+android_logger = "0.14"
+log = "0.4"
+tracing-log = "0.2"
 
 [build-dependencies]
 cc = "1"

crates/wzp-android/src/jni_bridge.rs

@@ -35,24 +35,17 @@ static INIT_LOGGING: Once = Once::new();
 /// Safe to call multiple times — only the first call takes effect.
 fn init_logging() {
     INIT_LOGGING.call_once(|| {
-        // Wrap in catch_unwind — sharded_slab allocation inside
-        // tracing_subscriber::registry() can crash on some Android
-        // devices if scudo malloc fails during early initialization.
+        // Use android_logger directly — tracing_subscriber::registry() allocates
+        // a sharded_slab which causes SIGSEGV on Android 16 MTE devices.
+        // android_logger is lightweight and doesn't trigger scudo crashes.
         let _ = std::panic::catch_unwind(|| {
-            use tracing_subscriber::layer::SubscriberExt;
-            use tracing_subscriber::util::SubscriberInitExt;
-            use tracing_subscriber::EnvFilter;
-            if let Ok(layer) = tracing_android::layer("wzp_android") {
-                // Filter: INFO for our crates, WARN for everything else.
-                // The jni crate emits VERBOSE logs for every method lookup
-                // (~10 lines per JNI call, 100+ calls/sec) which floods logcat
-                // and causes the system to kill the app.
-                let filter = EnvFilter::new("warn,wzp_android=info,wzp_proto=info,wzp_transport=info,wzp_codec=info,wzp_fec=info,wzp_crypto=info");
-                let _ = tracing_subscriber::registry()
-                    .with(layer)
-                    .with(filter)
-                    .try_init();
-            }
+            android_logger::init_once(
+                android_logger::Config::default()
+                    .with_max_level(log::LevelFilter::Info)
+                    .with_tag("wzp"),
+            );
+            // Bridge tracing → log so our tracing::info! macros work
+            let _ = tracing_log::LogTracer::init();
         });
     });
 }