Fix cuda.coop limitation preventing user-defined types when items_per_thread > 1 in block scan module.

[tpn]

Commits have been organized to ease review. Underlying "bug" was fixed by adding a missing lower_constant for the Complex type we use in our tests of user-defined types. _block_scan.py invariants and docstrings updated accordingly.

Additionally, now that there are no restrictions regarding items_per_thread > 1, I have changed the default from 1 to 4, such that we have a more optimal default out-of-the-box.

• New or existing tests cover these changes.
• The documentation is up to date with these changes.

This will close #4747 when merged.

[gevtushenko]

important: having default number of items per thread is a bit concerning. Even if there was a reasonable default, it's dictated by user data, not CUB. Since we don't know what user is about to pass (yet), I'd probably favor interface that forces user to spell this number explicitly.

[tpn]

I feel like this PR is the most sensible place to land that. I'll work on another commit now.

[tpn]

@gevtushenko see 6e59a37

[github-actions]

🟨 CI finished in 1h 24m: Pass: 83%/12 | Total: 1h 50m | Avg: 9m 12s | Max: 20m 00s

• 🟨 python: Pass: 83%/12 | Total: 1h 50m | Avg: 9m 12s | Max: 20m 00s

  🚨 jobs: Test cuda.cooperative 🚨
    🟩 Build cuda.cccl    Pass: 100%/2   | Total:  6m 31s | Avg:  3m 15s | Max:  3m 16s
    🟩 Build cuda.cooperative Pass: 100%/2   | Total:  6m 47s | Avg:  3m 23s | Max:  3m 28s
    🟩 Build cuda.parallel Pass: 100%/2   | Total: 16m 35s | Avg:  8m 17s | Max:  8m 23s
    🟩 Test cuda.cccl     Pass: 100%/2   | Total:  9m 04s | Avg:  4m 32s | Max:  4m 43s
    🔥 Test cuda.cooperative Pass:   0%/2   | Total: 37m 32s | Avg: 18m 46s | Max: 20m 00s
    🟩 Test cuda.parallel Pass: 100%/2   | Total: 34m 02s | Avg: 17m 01s | Max: 17m 04s
  🟨 cpu
    🟨 amd64              Pass:  83%/12  | Total:  1h 50m | Avg:  9m 12s | Max: 20m 00s
  🟨 ctk
    🟨 12.9               Pass:  83%/12  | Total:  1h 50m | Avg:  9m 12s | Max: 20m 00s
  🟨 cudacxx
    🟨 nvcc12.9           Pass:  83%/12  | Total:  1h 50m | Avg:  9m 12s | Max: 20m 00s
  🟨 cudacxx_family
    🟨 nvcc               Pass:  83%/12  | Total:  1h 50m | Avg:  9m 12s | Max: 20m 00s
  🟨 cxx
    🟨 GCC13              Pass:  83%/12  | Total:  1h 50m | Avg:  9m 12s | Max: 20m 00s
  🟨 cxx_family
    🟨 GCC                Pass:  83%/12  | Total:  1h 50m | Avg:  9m 12s | Max: 20m 00s
  🟨 gpu
    🟨 rtxa6000           Pass:  83%/12  | Total:  1h 50m | Avg:  9m 12s | Max: 20m 00s
  🟨 py_version
    🟨 3.10               Pass:  83%/6   | Total: 56m 20s | Avg:  9m 23s | Max: 20m 00s
    🟨 3.13               Pass:  83%/6   | Total: 54m 11s | Avg:  9m 01s | Max: 17m 32s
  

👃 Inspect Changes

Modifications in project?

	Project
	CCCL Infrastructure
	libcu++
	CUB
	Thrust
	CUDA Experimental
	stdpar
+/-	python
	CCCL C Parallel Library
	Catch2Helper

Modifications in project or dependencies?

	Project
	CCCL Infrastructure
	libcu++
	CUB
	Thrust
	CUDA Experimental
	stdpar
+/-	python
	CCCL C Parallel Library
	Catch2Helper

🏃‍ Runner counts (total jobs: 12)

#	Runner
6	linux-amd64-cpu16
6	linux-amd64-gpu-rtxa6000-latest-1

[github-actions]

🟩 CI finished in 27m 41s: Pass: 100%/12 | Total: 1h 51m | Avg: 9m 15s | Max: 19m 52s

• 🟩 python: Pass: 100%/12 | Total: 1h 51m | Avg: 9m 15s | Max: 19m 52s

  🟩 cpu
    🟩 amd64              Pass: 100%/12  | Total:  1h 51m | Avg:  9m 15s | Max: 19m 52s
  🟩 ctk
    🟩 12.9               Pass: 100%/12  | Total:  1h 51m | Avg:  9m 15s | Max: 19m 52s
  🟩 cudacxx
    🟩 nvcc12.9           Pass: 100%/12  | Total:  1h 51m | Avg:  9m 15s | Max: 19m 52s
  🟩 cudacxx_family
    🟩 nvcc               Pass: 100%/12  | Total:  1h 51m | Avg:  9m 15s | Max: 19m 52s
  🟩 cxx
    🟩 GCC13              Pass: 100%/12  | Total:  1h 51m | Avg:  9m 15s | Max: 19m 52s
  🟩 cxx_family
    🟩 GCC                Pass: 100%/12  | Total:  1h 51m | Avg:  9m 15s | Max: 19m 52s
  🟩 gpu
    🟩 rtxa6000           Pass: 100%/12  | Total:  1h 51m | Avg:  9m 15s | Max: 19m 52s
  🟩 jobs
    🟩 Build cuda.cccl    Pass: 100%/2   | Total:  6m 56s | Avg:  3m 28s | Max:  3m 35s
    🟩 Build cuda.cooperative Pass: 100%/2   | Total:  6m 56s | Avg:  3m 28s | Max:  3m 35s
    🟩 Build cuda.parallel Pass: 100%/2   | Total: 18m 31s | Avg:  9m 15s | Max:  9m 19s
    🟩 Test cuda.cccl     Pass: 100%/2   | Total:  8m 59s | Avg:  4m 29s | Max:  4m 45s
    🟩 Test cuda.cooperative Pass: 100%/2   | Total: 37m 15s | Avg: 18m 37s | Max: 19m 52s
    🟩 Test cuda.parallel Pass: 100%/2   | Total: 32m 34s | Avg: 16m 17s | Max: 17m 10s
  🟩 py_version
    🟩 3.10               Pass: 100%/6   | Total: 55m 47s | Avg:  9m 17s | Max: 17m 23s
    🟩 3.13               Pass: 100%/6   | Total: 55m 24s | Avg:  9m 14s | Max: 19m 52s
  

👃 Inspect Changes

Modifications in project?

	Project
	CCCL Infrastructure
	libcu++
	CUB
	Thrust
	CUDA Experimental
	stdpar
+/-	python
	CCCL C Parallel Library
	Catch2Helper

Modifications in project or dependencies?

	Project
	CCCL Infrastructure
	libcu++
	CUB
	Thrust
	CUDA Experimental
	stdpar
+/-	python
	CCCL C Parallel Library
	Catch2Helper

🏃‍ Runner counts (total jobs: 12)

#	Runner
6	linux-amd64-cpu16
6	linux-amd64-gpu-rtxa6000-latest-1