Add online speech denoiser for GTCRN and examples (#3324)

This PR adds online (streaming) speech denoising support for GTCRN alongside existing DPDFNet support, and updates multi-language bindings/examples to expose and validate the new capability.

java-api-examples/StreamingSpeechEnhancementDpdfNet.java

@@ -0,0 +1,64 @@
+// Copyright 2026 Xiaomi Corporation
+//
+// This file shows how to use streaming DPDFNet speech enhancement models in
+// sherpa-onnx.
+//
+// Download DPDFNet models from either:
+// https://github.com/k2-fsa/sherpa-onnx/releases/tag/speech-enhancement-models
+// https://huggingface.co/Ceva-IP/DPDFNet
+
+import com.k2fsa.sherpa.onnx.*;
+
+public class StreamingSpeechEnhancementDpdfNet {
+  private static void appendSamples(java.util.ArrayList<Float> dst, float[] src) {
+    for (float v : src) {
+      dst.add(v);
+    }
+  }
+
+  private static float[] toFloatArray(java.util.ArrayList<Float> src) {
+    float[] ans = new float[src.size()];
+    for (int i = 0; i != src.size(); ++i) {
+      ans[i] = src.get(i);
+    }
+    return ans;
+  }
+
+  public static void main(String[] args) {
+    String model = "./dpdfnet_baseline.onnx";
+
+    OfflineSpeechDenoiserModelConfig modelConfig =
+        OfflineSpeechDenoiserModelConfig.builder()
+            .setNumThreads(1)
+            .setDebug(true)
+            .setProvider("cpu")
+            .setDpdfnet(
+                OfflineSpeechDenoiserDpdfNetModelConfig.builder().setModel(model).build())
+            .build();
+
+    OnlineSpeechDenoiserConfig config =
+        OnlineSpeechDenoiserConfig.builder().setModel(modelConfig).build();
+
+    OnlineSpeechDenoiser speechDenoiser = new OnlineSpeechDenoiser(config);
+
+    WaveReader reader = new WaveReader("./inp_16k.wav");
+    int frameShift = speechDenoiser.getFrameShiftInSamples();
+    java.util.ArrayList<Float> output = new java.util.ArrayList<>();
+
+    float[] samples = reader.getSamples();
+    for (int start = 0; start < samples.length; start += frameShift) {
+      int end = Math.min(start + frameShift, samples.length);
+      float[] chunk = java.util.Arrays.copyOfRange(samples, start, end);
+      DenoisedAudio denoised = speechDenoiser.run(chunk, reader.getSampleRate());
+      appendSamples(output, denoised.getSamples());
+    }
+
+    DenoisedAudio denoised = speechDenoiser.flush();
+    appendSamples(output, denoised.getSamples());
+    String outFilename = "enhanced-online-dpdfnet.wav";
+    WaveWriter.write(outFilename, toFloatArray(output), speechDenoiser.getSampleRate());
+    System.out.printf("Saved to %s\n", outFilename);
+
+    speechDenoiser.release();
+  }
+}