refactor(examples): wrap logic in main() functions

[OfficialSerge]

Description

This PR is the second half of the response to the following TODO item in cuda_core/tests/example_tests/utils.py.

see part 1

# TODO: Refactor the examples to give them a common callable `main()` to avoid needing to use exec here?
exec(script, env if env else {})

Key Changes

• refactor any remaining non-module examples into Python modules.
• os.path.join(cuda_path, "include") refactor to Path(cuda_path) / "include" in thread_block_cluster.py.
• os.path.isdir(cuda_include) refactor to cuda_include.is_dir() in thread_block_cluster.py.
• some indenting of the code blocks, they now have leading white space technically, this shouldn't break anything.
• tried keeping everything else the same.

Next Steps

• consider using contextlib.closing for Stream and Buffer objects, current failed assertions will halt the execution of the example before .close() is called, thus cleanup is now the garbage collector's problem, this is non-deterministic though; since the Python's garbage collector is lazy, cleanup isn't immediate, not the end of the world but Streams or Buffers linger in the GPU drivers. I imagine many failed tests due to assertion errors may cause unneeded strain on GPU resources.

Let me know if this sounds good (seems okay but not critical imo).

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.