[FEA]: [CUB] Add a env-based overload of the device segmented reductions primitives #6673

Is this a duplicate?

• I confirmed there appear to be no duplicate issues for this request and that I agree to the Code of Conduct

Area

CUB

Is your feature request related to a problem? Please describe.

Citing @gevtushenko in this discussion to describe the need:

we are actually lacking env-based overload on the device segmented reduce taking offse iterators. Having environment-based version there would unblock significant optimization opportunities. Analogous to requirements API, we considered adding guarantees API with which user would be able to tell us about, say, max segment size. On the implementaion side, we'd be able to choose optimal kernel implementation statically. Similar idea applies to segmented sort.

Note: this #6620 was also opened regarding requirement API.

Corresponding PR: #6674

Describe the solution you'd like

env-based overloads of the device segmented reductions primitives are desired.

Describe alternatives you've considered

No response

Additional context

No response