[Bugfix][NVSHMEM] Support v2 state in legacy IBGDA kernels

[XFDG]

Summary

Port NVSHMEM v2 IBGDA RC indexing support into DeepEP's legacy kernels.

Fix

• Preserve the existing RC mapping for NVSHMEM v1 state.
• Use the v2 PE/NPES indexing scheme for the newer state layout.
• Define __CUDACC_RDC__ when NVSHMEM is enabled so device symbols use the correct extern declarations.

Testing

• Full SM100 extension build, NVSHMEM device-link, and shared-library link passed with CUDA 13.1, NVSHMEM 3.5.21, and isolated NCCL 2.32.3.
• csrc/kernels/legacy/internode.cu and internode_ll.cu also compiled for SM100 with NVSHMEM 3.4.5.
• An extension import smoke test failed in the available Torch environment with ImportError: unsupported scalarType; therefore runtime behavior is not claimed as verified.
• Cross-node IBGDA runtime testing was not available on the single 8-GPU node.

Fixes #764

[XFDG]

Validation update for head 2ba5e7d: on B200/SM100 with CUDA 13.1, NVSHMEM 3.5.21 and NCCL 2.32.3, the extension completes all 9 build/link steps. Loading the built deep_ep._C artifact and importing deep_ep (including init_jit) now succeeds with the matching NCCL 2.32.3 directory first in LD_LIBRARY_PATH. Without that setting, the process loads system NCCL 2.29.2 and the package correctly rejects the version mismatch. This confirms the current import smoke in the aligned environment; I have not run a multi-node IBGDA runtime test.

[XFDG]

Closing this PR because the current main branch removed the affected legacy IBGDA implementation in DeepEP V2.5 (#763). The remaining NVSHMEM compilation guard also has no current source/build consumer, so restoring the deleted legacy code merely to resolve the conflict would not be appropriate. This is an obsolete fix, not a claim that this PR was merged. Historical validation results remain in the PR for reference. Thanks for reviewing.