[fix] fix plugin te_general_grouped test bug (#81)

## Summary

Replace the grouped GEMM plugin test import from
`transformer_engine_torch_nv` to `transformer_engine_torch`.

## Why

The test should load the standard Transformer Engine torch extension
module. Importing `transformer_engine_torch_nv` can fail because that is
not the expected module name in this project.

## Impact

This fixes startup/import failure for
`transformer_engine/plugin/tests/test_te_general_grouped.py` before the
grouped GEMM checks run.

## Validation

- `git diff --check origin/main...HEAD`
- `python3 -c 'import ast, pathlib; path =
pathlib.Path("transformer_engine/plugin/tests/test_te_general_grouped.py");
ast.parse(path.read_text())'`

Full test execution was not run locally because it requires the project
runtime/CUDA environment.

Author: lihongyang1990

transformer_engine/plugin/tests/test_te_general_grouped.py

@@ -22,7 +22,7 @@ def test_grouped_gemm_equivalence(self, grad, has_bias, has_pre_gelu, single_out
             "\n test te_general_grouped_gemm"
             f" grad:{grad} has_bias:{has_bias},has_pre_gelu:{has_pre_gelu},single_output:{single_output}"
         )
-        import transformer_engine_torch_nv as tex
+        import transformer_engine_torch as tex
 
         num_gemms = 2
         m, k, n = 128, 32, 64