Skip RWKV test in past CI

[ydshieh]

What does this PR do?

In the past CI with torch 13.1 (and older), RWKV tests fail.

• The first failure is

  test_model_parallelism
  
  (line 114)  RuntimeError: CUDA error: CUBLAS_STATUS_EXECUTION_FAILED when calling cublasSgemm( handle, opa, opb, m, n, k,   &alpha, a, lda, b, ldb, &beta, c, ldc)

  • this affects 1x more subsequential tests
  • skip this particular test will reduce a lot the number of failures.

• The docker env. of that past CI uses a base docker image shipped with cuda 11.6

• In a docker image with cuda 11.8 but installing torch==1.13+cu116, no failure.

Let's not take too much time on identifying what exactly cause the failure here, and just skip RWKV tests on past CI with torch < 2.0.

The goal is just to have a clean past CI report.

[sgugger]

Works for me!