Make SetSubData take size in uint8_t not uint32_t.

Also change data pointer to uint8_t*.
Add a test for a non-aligned SetSubData().
Implemented on all backends.

examples/ComputeBoids.cpp

@@ -88,8 +88,8 @@ void initBuffers() {
             .GetResult();
 
         particleBuffers[i].SetSubData(0,
-            sizeof(Particle) * kNumParticles / sizeof(uint32_t),
-            reinterpret_cast<uint32_t*>(initialParticles.data()));
+            sizeof(Particle) * kNumParticles,
+            reinterpret_cast<uint8_t*>(initialParticles.data()));
     }
 }