Add stride for mapped direct array parameters

[jeffamstutz]

While not common, it has come up in practice that direct array parameters may want to allow writing sparse data arrays internal to the 3D rendering engine. This specifically came up in the Cycles implementation where VEC3 data is 16-byte aligned -- thus a staging buffer is required for the current formulation of direct parameter arrays.

The remedy is quite simple -- have a stride out-parameter that implementations use to calculate the offset into the memory for each element. This would look like:

ANARIGeometry geom = anariNewGeometry(device, "triangle");
size_t outStride = 0;
void *vertices = anariMapParameterArray1D(device, geom, "vertex.position", ANARI_FLOAT32_VEC3, numVertices, &outStride);
for (int i = 0; i < numVertices, i++) {
  float3 *out = vertices + i * outStride;
  *out = inVertices[i];
} 
// ...