Add case-weighted normalisation statistics (each training trajectory counts equally)

data/stats_case_weighted.yaml

@@ -0,0 +1,39 @@
+# TurtLES normalisation statistics: train split(s), 320 cases,
+# all 400 frames and all points per case, weighting = cases.
+# delta = x(t+1) - x(t) between consecutive saved frames (399 per case).
+mean:
+  p: 0.9915
+  rho: 0.99195
+  u: 0.19144
+  v: 0.00053878
+  w: 8.3904e-06
+rms:
+  p: 0.99165
+  rho: 0.99206
+  u: 0.21815
+  v: 0.069433
+  w: 0.030743
+std:
+  p: 0.016981
+  rho: 0.014865
+  u: 0.10459
+  v: 0.069431
+  w: 0.030743
+mean_delta:
+  p: -5.7213e-08
+  rho: -4.8532e-08
+  u: 6.1494e-08
+  v: -2.8449e-07
+  w: -1.6985e-08
+rms_delta:
+  p: 0.0010256
+  rho: 0.0010874
+  u: 0.0090419
+  v: 0.010083
+  w: 0.0097833
+std_delta:
+  p: 0.0010256
+  rho: 0.0010874
+  u: 0.0090419
+  v: 0.010083
+  w: 0.0097833