Make the new CUB tuning API public

[bernhardmgruber]

For each CUB algorithm, if the tuning API has been implemented internally, we are ready to make it publicly accessible:

• Enable public access to tuning properties for CUB environment API overloads. Add a test for each CUB algo overload specifying a custom tuning.
• Add documentation and examples on how the public tuning API works
• Deprecated all public CUB dispatchers (dropped at the next major release)