python: bind PixelHistogramCUDA

PixelHistogramCUDA_<input>_<storage> for uint16, float32 and float64
frames and the six storage types of PixelHistogram, behind an
aare.PixelHistogramCUDA() factory. Frames are never converted: a wrong
dtype raises TypeError, so a registered buffer is always the one copied.
The GIL is released during GPU work.

- C++: fill_histogram() and submit_batch() take const views, so
  read-only arrays (np.memmap mode="r") work; register_output_buffer()
  pins the destination of values(out).
- Python: register_input_buffer()/register_output_buffer() keep their
  array alive; values(out=...) writes into a given array. With a pinned
  out, reading back a 640 MB histogram takes 28 ms instead of 230 ms.
- Tests in test_PixelHistogramCUDA.py; PixelHistogramCUDA.ipynb compares
  CPU and GPU on 100k Moench frames (identical, ~13x faster on the GPU).

python/src/cuda_bindings.cu

@@ -4,11 +4,13 @@
 // the ClusterVector and Cluster types it exposes in its return values, so
 // the module is self-contained — users can call steal_clusters() and get
 // back a usable ClusterVector without _aare needing to be imported first.
+// Also registers PixelHistogramCUDA.
 
 #include "bind_Cluster.hpp"
 #include "bind_ClusterFinderCUDA.hpp"
 #include "bind_ClusterFinderCUDAGraph.hpp"
 #include "bind_ClusterVector.hpp"
+#include "bind_PixelHistogramCUDA.hpp"
 
 #include <pybind11/pybind11.h>
 
@@ -83,6 +85,9 @@ PYBIND11_MODULE(_aare_cuda, m) {
     DEFINE_BINDINGS_CLUSTERFINDER_CUDA_GRAPH(int, 9, 9, uint16_t, i);
     DEFINE_BINDINGS_CLUSTERFINDER_CUDA_GRAPH(double, 9, 9, uint16_t, d);
     DEFINE_BINDINGS_CLUSTERFINDER_CUDA_GRAPH(float, 9, 9, uint16_t, f);
+
+    // Pixel histograms: PixelHistogramCUDA_<input>_<storage>
+    aare::define_PixelHistogramCUDA_bindings(m);
 }
 
 #undef DEFINE_CUDA_CLUSTER_TYPES