Add the DeviceSelect::FlaggedIf algorithm

[gonidelis]

Description

This adds the cub::DeviceSelect::FlaggedIf algorithm which combines the predicate selection approach of cub::DeviceSelect::If and the flag approach of cub::DeviceSelect::Flagged.

This fixes #1409.

Checklist

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

[elstehle]

Thank you for picking this up!

Overall. this looks good already. I think we want to add documentation and examples for the newly introduced interfaces and extend test coverage a bit more.

[gonidelis]

Thank you for picking this up!

Overall. this looks good already. I think we want to add documentation and examples for the newly introduced interfaces and extend test coverage a bit more.

Working on the tests rn - that's why I submitted it as a draft. Well, on the documentation too... :|

[elstehle]

Thanks a lot for the great work! Looks good, just a couple of minor suggestions.

[gonidelis]

@elstehle I fixed everything except the get_reference() portion of it in the unit test, which I want to isolate in a separate commit. Thanks for the feedback!

ps: Need to also extract the in-place API example in an separate executable test.

[gevtushenko]

Looks good! Just to make sure that in-place overload compiles, I'd like to see its snippet extracted into API test.

[elstehle]

Just two minor nitpicks remaining, otherwise good to go 🚀