Initial commit: E4B-MarkBase model integration with passing tests

- E4B-MarkBase model (42 layers, 4.4GB) loaded successfully
- All Phase 1-6 tests passed (model loading, forward pass, vision/audio towers, token generation, performance)
- All stress tests passed (5/5 in 127.6s)
  - Concurrent inference
  - Memory stress (67.5 tok/s, 0 NaN)
  - Continuous generation
  - Batch processing
  - Long-running stability
- Swift Metal inference engine with multimodal support

Sources/MarkBaseServer/APIServer.swift

@@ -0,0 +1,31 @@
+import MarkBase
+
+/// Entry point for MarkBase API server
+/// Usage: swift run G12BServer [model_dir] [port] [model_id] [--benchmark]
+@main
+public struct ServerMain {
+    public static func main() async throws {
+        let args = CommandLine.arguments
+        
+        // Check for benchmark mode
+        if args.contains("--benchmark") {
+            let modelDir = args.count > 2 ? args[1] : "./model"
+            let modelName = args.count > 3 ? args[2] : "markbase"
+            
+            var benchmark = PerformanceBenchmark(modelDir: modelDir, modelName: modelName)
+            try await benchmark.run()
+            return
+        }
+        
+        let modelDir = args.count > 1 ? args[1] : "./model"
+        let port = args.count > 2 ? Int(args[2]) ?? 8080 : 8080
+        let modelId = args.count > 3 ? args[3] : "markbase-12b"
+        
+        print("\n╔═════════════════════════════════════════════╗")
+        print("║      MarkBase API Server                    ║")
+        print("╚═════════════════════════════════════════════╝\n")
+        
+        let server = try MarkBaseServer(modelDir: modelDir, modelId: modelId)
+        try await server.start(port: port)
+    }
+}