add OpenBLAS detection and modify tests codes

[katsu560]

I revised src/CMakeLists.txt for fixing indents and commands for Haiku and adding OpenBLAS detection.
OpenBLAS detection codes in whisper CMakeLists.txt isn't enough. That detects only OpenBLAS library.
Thus, I added codes for detecting OpenBLAS header based on caffe's FindOpenBLAS.cmake.

I also revised tests/CMakeLists.txt for adding OpenBLAS detection and support x86 system w/o FMA instructions.
And some tests code are modified.

test-vec1.c:
The fmadd codes are changed to without FMA codes on no FMA support system.

test-grad0.c test-mul-mat0.c:
The loop number is changeable by environment variable GGML_NLOOP or command line option.
Because, 1000 or 500 loop are too much to finish loop in a time.

test2.c:
n_threads is changeable by environment variable GGML_NTHREADS or command line option.
Because my 4-logical core cpu pc can't do test2 with 8 threads setting.

After adapted my modifications, the results of make test is as follows on my old SandyBridge Linux pc.

$ GGML_NLOOP=1 GGML_NTHREADS=4 make test
Running tests...
Test project /home/user/github/gpt/ggml/build
      Start  1: test-vec0
 1/10 Test  #1: test-vec0 ........................   Passed    4.01 sec
      Start  2: test-vec1
 2/10 Test  #2: test-vec1 ........................   Passed   11.79 sec
      Start  3: test-grad0
 3/10 Test  #3: test-grad0 .......................   Passed  147.97 sec
      Start  4: test-mul-mat0
 4/10 Test  #4: test-mul-mat0 ....................   Passed   35.14 sec
      Start  5: test-mul-mat2
 5/10 Test  #5: test-mul-mat2 ....................   Passed    2.49 sec
      Start  6: test0
 6/10 Test  #6: test0 ............................   Passed    2.46 sec
      Start  7: test1
 7/10 Test  #7: test1 ............................   Passed   25.39 sec
      Start  8: test2
 8/10 Test  #8: test2 ............................   Passed   12.83 sec
      Start  9: test3
 9/10 Test  #9: test3 ............................   Passed   12.71 sec
      Start 10: test-svd0
10/10 Test #10: test-svd0 ........................   Passed    2.36 sec

100% tests passed, 0 tests failed out of 10

Total Test time (real) = 257.22 sec