[cuBLASLt] relax addmm cuBLASLt constraint (#153675)

`beta == 1.0` doesn't seem to be required anymore

https://github.com/pytorch/pytorch/issues/153590

`self.dim() == 1` restriction seems to still hold but not sure if that's due to a lack of handling on the PyTorch side or the cuBLASLt side, will investigate

Pull Request resolved: https://github.com/pytorch/pytorch/pull/153675
Approved by: https://github.com/Skylion007

aten/src/ATen/native/cuda/Blas.cpp

@@ -369,7 +369,7 @@ Tensor& addmm_out_cuda_impl(Tensor& result, const Tensor& self, const Tensor& ma
     // leading dim >> rows when they are sliced from a large tensor
     // see fbcode/caffe2/test/test_linalg.py:test_corner_cases_of_cublasltmatmul
     if (!disable_addmm_cuda_lt_final) {
-      useLtInterface = beta.toComplexDouble() == 1.0 && self.dim() == 1 &&
+      useLtInterface = self.dim() == 1 &&
           result.dim() == 2 && self.sizes()[0] == mat2_sizes[1] &&
           self.is_contiguous() && result.is_contiguous() &&
 #ifdef USE_ROCM