Avoid integer overflow when computing the sample size

While it seems to be harmless in this particular case, it still prevents
testing this code with UBSAN by triggering it here, so check that
multiplication doesn't overflow.

src/unix/sound.cpp

@@ -673,10 +673,16 @@ bool wxSound::LoadWAV(const void* data_, size_t length, bool copyData)
         waveformat.ulSamplesPerSec * waveformat.uiBlockAlign)
         return false;
 
-    // We divide by this number below to obtain the number of samples, so it
+    // We divide by the sample size below to obtain the number of samples, so
-    // definitely can't be 0.
+    // it definitely can't be 0. Also take care to avoid integer overflow when
-    wxUint32 const sampleSize =
+    // computing it.
-        waveformat.uiChannels * waveformat.uiBitsPerSample / 8;
+    unsigned tmp = waveformat.uiChannels;
+    if (tmp >= 0x10000)
+        return false;
+
+    tmp *= waveformat.uiBitsPerSample;
+
+    wxUint32 const sampleSize = tmp / 8;
     if (!sampleSize)
         return false;