updated all ver to 2.4.0, tweaked CHANGELOG

Author: ahbarnett

CHANGELOG

@@ -1,8 +1,10 @@
 List of features / changes made / release notes, in reverse chronological order.
 If not stated, FINUFFT is assumed (cuFINUFFT <=1.3 is listed separately).
 
-Master, using release name V 2.4.0 (4/30/25)
+V 2.4.0 (5/27/25)
 
+* slight accuracy improvements for small N by use of ceil in nf choice
+  (Barbone & Barnett a0bde26).
 * PR #662 (Barbone):
   Direct calc (dirft) templated, norms computed by hand in tests, to remove
   finufft dependency for cufinufft math tests.
@@ -24,7 +26,7 @@ Master, using release name V 2.4.0 (4/30/25)
 * fixed empty-opts bug in MATLAB/Octave for {cu}finufft, PR 666 (Barnett).
 * MATLAB/Octave unified fullmathtest.m for CPU and/or GPU (Barnett).
 * MATLAB GPU interface (MEX CUDA) via gpuArrays, PR 634 (mostly Libin Lu;
-  docs/examples/tests/interfaces by Barnett).
+  docs/examples/tests/interfaces by Barnett; sped up by Fortunato).
 * Tweaked choice of upsampfac to use a density based heuristic for type 1 and 2
   in the CPU library. This gives a significant speedup for some cases.
   For now we assume density=1.0 (to revisit).  Barbone. PR #638 & 652.

python/cufinufft/cufinufft/__init__.py

@@ -8,4 +8,4 @@
            "nufft3d1", "nufft3d2",
            "Plan"]
 
-__version__ = '2.4.0beta1'
+__version__ = '2.4.0'

python/finufft/finufft/__init__.py

@@ -17,4 +17,4 @@
 from finufft._interfaces import nufft2d1,nufft2d2,nufft2d3
 from finufft._interfaces import nufft3d1,nufft3d2,nufft3d3
 
-__version__ = '2.4.0-rc1'
+__version__ = '2.4.0'