Fixes for Python 3.7 docs environment

[shwina]

Description

Docs builds are currently failing because cuda.parallel is not importable on Python 3.7 environments (which our docs build environment is stuck on). This PR fixes that by making cuda.parallel importable with Python 3.7.

Additional Context

The Sphinx autodoc extension requires modules to be importable.

Checklist

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

[github-actions]

🟩 CI finished in 24m 37s: Pass: 100%/1 | Total: 24m 37s | Avg: 24m 37s | Max: 24m 37s

• 🟩 python: Pass: 100%/1 | Total: 24m 37s | Avg: 24m 37s | Max: 24m 37s

  🟩 cpu
    🟩 amd64              Pass: 100%/1   | Total: 24m 37s | Avg: 24m 37s | Max: 24m 37s
  🟩 ctk
    🟩 12.6               Pass: 100%/1   | Total: 24m 37s | Avg: 24m 37s | Max: 24m 37s
  🟩 cudacxx
    🟩 nvcc12.6           Pass: 100%/1   | Total: 24m 37s | Avg: 24m 37s | Max: 24m 37s
  🟩 cudacxx_family
    🟩 nvcc               Pass: 100%/1   | Total: 24m 37s | Avg: 24m 37s | Max: 24m 37s
  🟩 cxx
    🟩 GCC13              Pass: 100%/1   | Total: 24m 37s | Avg: 24m 37s | Max: 24m 37s
  🟩 cxx_family
    🟩 GCC                Pass: 100%/1   | Total: 24m 37s | Avg: 24m 37s | Max: 24m 37s
  🟩 gpu
    🟩 v100               Pass: 100%/1   | Total: 24m 37s | Avg: 24m 37s | Max: 24m 37s
  🟩 jobs
    🟩 Test               Pass: 100%/1   | Total: 24m 37s | Avg: 24m 37s | Max: 24m 37s
  

👃 Inspect Changes

Modifications in project?

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

Modifications in project or dependencies?

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

🏃‍ Runner counts (total jobs: 1)

#	Runner
1	linux-amd64-gpu-v100-latest-1