[CuTe] Handle Underscore entries in zip2_by

[XFDG]

Summary

• preserve an entire layout mode in REST when its zip2_by guide entry is Underscore
• let shape and stride callers provide the correct BLOCK identity (1 and 0, respectively)
• cover both hierarchical and scalar kept modes in the CuTe core unit tests

Testing

• B200 (SM100a), CUDA 13.1: hierarchical-mode baseline produced the wrong layout; fixed repro compiled and returned 0 with the expected layout
• B200 (SM100a), CUDA 13.1: scalar-mode baseline failed to compile in zip2_by; fixed repro compiled and returned 0 with the expected layout
• full cutlass_test_unit_cute_core target was not run because CMake's GoogleTest FetchContent cannot reach GitHub from the GPU node

Contribution disclosure

AI assistance was used for repository research, implementation, and running the checks above.

Fixes #3514