Add cub::detail::BlockLoadToShared (#5780)

* Design BlockLoadToShared

* Add Invalidate member function
  Stop invalidating in destructor

* Peel both ends

* Add span of const overload for usability

* Use async-group based sync for LDGSTS

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

Author: pauleonix