Improve performance of Euclidean distance transform

[grlee77]

This PR gives a further 2-4x speedup to the 2D and 3D distance transform functions by creating dedicated elementwise CUDA kernels for the pre- and post-processing steps. These include, for example, packing and upacking the 3D coordinates into a single integer. This resolves various "TODO" comments that were in the source, but I don't think we have an issue open for this.

Aside from the performance benefit there is also a substantial reduction in memory usage for the default case of return_distance=True, return_indices=False as in this case the explicit arrays of indices never need to be created.

Benchmarks

Implementation from release 22.08.01

shape	return dist.	return ind.	SciPy	ITK	cuCIM	accel. (CuPy vs SciPy)	accel. (CuPy vs ITK)
256, 256	False	True	0.00379	----	0.00081	4.66	----
256, 256	True	False	0.00428	0.00383	0.00102	4.21	3.76
256, 256	True	True	0.00429	----	0.00107	4.02	----
512, 512	False	True	0.01724	----	0.00091	18.97	----
512, 512	True	False	0.01866	0.00788	0.00104	17.94	7.58
512, 512	True	True	0.01876	----	0.00107	17.48	----
2048, 2048	False	True	0.35053	----	0.00256	137.03	----
2048, 2048	True	False	0.39468	0.08220	0.00288	137.02	28.54
2048, 2048	True	True	0.39356	----	0.00296	133.04	----
4096, 4096	False	True	1.57430	----	0.00736	213.85	----
4096, 4096	True	False	1.79513	0.54968	0.00858	209.34	64.10
4096, 4096	True	True	1.81570	----	0.00889	204.25	----
32, 32, 32	False	True	0.00319	----	0.00069	4.62	----
32, 32, 32	True	False	0.00356	0.00481	0.00096	3.71	5.01
32, 32, 32	True	True	0.00360	----	0.00103	3.49	----
128, 128, 128	False	True	0.27901	----	0.00131	213.20	----
128, 128, 128	True	False	0.31013	0.04504	0.00156	199.32	28.94
128, 128, 128	True	True	0.31289	----	0.00163	192.07	----
256, 256, 256	False	True	3.02435	----	0.00696	434.25	----
256, 256, 256	True	False	3.34077	0.33528	0.00855	390.57	39.20
256, 256, 256	True	True	3.32920	----	0.00915	363.89	----
384, 384, 384	False	True	10.68973	----	0.02284	468.07	----
384, 384, 384	True	False	11.73340	1.16172	0.02809	417.65	41.35
384, 384, 384	True	True	11.68002	----	0.03007	388.40	----

Proposed Implementation

shape	return dist.	return ind.	SciPy	ITK	cuCIM	accel. (CuPy vs SciPy)	accel. (CuPy vs ITK)
256, 256	False	True	0.00385	N/A	0.00021	18.49	N/A
256, 256	True	False	0.00433	0.00370	0.00021	20.84	17.83
256, 256	True	True	0.00436	N/A	0.00022	19.55	N/A
512, 512	False	True	0.01724	N/A	0.00030	58.44	N/A
512, 512	True	False	0.01893	0.00790	0.00029	66.42	27.73
512, 512	True	True	0.01867	N/A	0.00030	62.65	N/A
2048, 2048	False	True	0.33893	N/A	0.00154	220.25	N/A
2048, 2048	True	False	0.40912	0.08219	0.00154	266.11	53.46
2048, 2048	True	True	0.39056	N/A	0.00162	241.76	N/A
4096, 4096	False	True	1.58069	N/A	0.00418	377.92	N/A
4096, 4096	True	False	1.81860	0.55152	0.00418	435.36	132.03
4096, 4096	True	True	1.80671	N/A	0.00448	402.85	N/A
32, 32, 32	False	True	0.00319	N/A	0.00017	19.05	N/A
32, 32, 32	True	False	0.00359	0.00514	0.00011	33.25	47.68
32, 32, 32	True	True	0.00355	N/A	0.00019	18.37	N/A
128, 128, 128	False	True	0.27217	N/A	0.00059	458.70	N/A
128, 128, 128	True	False	0.31409	0.03384	0.00050	633.12	68.21
128, 128, 128	True	True	0.31207	N/A	0.00066	474.26	N/A
256, 256, 256	False	True	3.01718	N/A	0.00304	993.26	N/A
256, 256, 256	True	False	3.32660	0.28979	0.00235	1418.07	123.53
256, 256, 256	True	True	3.31874	N/A	0.00352	942.49	N/A
384, 384, 384	False	True	10.67328	N/A	0.01000	1067.34	N/A
384, 384, 384	True	False	11.69219	1.11274	0.00771	1517.46	144.42
384, 384, 384	True	True	11.76073	N/A	0.01161	1013.03	N/A

[gigony]

Thanks @grlee77 , it looks good to me!

[grlee77]

@gpucibot merge