cuTile-specific launch macro

`@cuda backend=cuTile ...` is slightly verbose, and gets repetitive, so I've been using a `@cutile ...` macro that automatically inserts the backend argument into the kernel launch expression. This is just a three-line definition, but would be useful as a centralized and exported interface for launching tile kernels.

Locate the existing kernel launch macro and the module's exported-interface declarations. Define the centralized @cutile interface so it inserts the cuTile backend argument, then verify that the macro is exported and expands correctly for the launch expression described in the issue.

Written by the indexing model from the issue text.