a major code release of rocblas

[tingxingdong]

Summary of proposed changes:

• cleaning: remove/rename the files with clblas, ablas,
• rocBLAS:
• provide cublas-like / batched API (four BLAS-1, GEMV, HEMV, GEMM)
• four BLAS-1 routines implementations: scal, swap, copy, dot
• auxilary routines: get_(set) vector, get(set)_matrix.
• Gtest accuracy testing:
• BLAS-1 routine scal testing
• BLAS-3 GEMM testing
• Benchmarking:
  • support scal, symv, dot
  • benchmarking client: accept command line
• Common operation
  • timing function, random generator, device query...
  • cblas compatible interface

[tingxingdong]

revision based on feedbacks:
1)get_wtime --> get_time_ms;
2)gflops --> gflop_count;
3)fix the indentation
4)add comments when using feature of c++11 explicitly
5) dir tests/ --> gtest/, files now ends with "_gtest.cpp"
6) remove get(set)_vector, replace them with hipMemcy
7) pointer_type --> pointer_location
8) fix typo: auxilary --> auxiliary
9) macro name CHECK_ERROR --> CHECK_HIP_ERROR, when using HIP API (e.g hipMallco, hipMemcpy)
10) delete unused files main.cpp. delete the unused expert interface related files.