Method cumsum

tests/python/relax/test_transform_static_plan_block_memory.py:1582–1583

(var_A: T.handle, var_A_1: T.handle, var_exclusive_scan_thrust: T.handle)

    class Module:
        @T.prim_func(private=True, s_tir=True)
        def cumsum(var_A: T.handle, var_A_1: T.handle, var_exclusive_scan_thrust: T.handle):
            T.evaluate(0)

        @R.function
        def main(probs: R.Tensor(("batch_size", "vocab_size"), dtype="float32")) -> R.Tensor(