Switch to Processes for voronoi cell volume calculation

Ref: Speed up dcf calculation #169

Scipy.spatial Qhull does not seem to work in parallel in multiple threads, using processes makes it~4x faster on my notebook.
Would still prefer a gpu version.

Author: fzimmermann89

src/mrpro/data/_DcfData.py

@@ -15,7 +15,7 @@
 from __future__ import annotations
 
 import dataclasses
-from concurrent.futures import ThreadPoolExecutor
+from concurrent.futures import ProcessPoolExecutor
 from functools import reduce
 from itertools import product
 
@@ -30,6 +30,10 @@
 UNIQUE_ROUNDING_DECIMALS = 15
 
 
+def _volume(v):
+    return ConvexHull(v).volume
+
+
 @dataclasses.dataclass(slots=True, frozen=False)
 class DcfData:
     """Density compensation data (DcfData) class."""
@@ -114,8 +118,9 @@ def _dcf_2d3d_voronoi(traj: torch.Tensor) -> torch.Tensor:
         regions = [vdiagram.regions[r] for r in vdiagram.point_region[: -len(corner_points)]]  # Ignore corner points
         vertices = [vdiagram.vertices[region] for region in regions]
 
-        # Calculate volume/area of voronoi cells using threads (ConvexHull is thread safe and drops the GIL)
-        future = ThreadPoolExecutor(max_workers=torch.get_num_threads()).map(lambda v: ConvexHull(v).volume, vertices)
+        # Calculate volume/area of voronoi cells using processes, as this is a very time-consuming operation
+        # and ConvexHull is singlethreaded and does not seem to drop the GIL
+        future = ProcessPoolExecutor(max_workers=torch.get_num_threads()).map(_volume, vertices, chunksize=100)
         dcf = np.array(list(future))
 
         # Get outliers (i.e. voronoi cell which are unbound) and set them to a reasonable value