benchmark-matmult : do not use integer abs() on a float

[cebtenzzre]

Fixes this clang warning, which I only saw when I was using clang 3.7:

examples/benchmark/benchmark-matmult.cpp:253:23: warning: using integer absolute value function 'abs' when argument is of
      floating point type [-Wabsolute-value]
        float delta = abs(sum_of_Q4_result - sum_of_F32_reference);
                      ^
examples/benchmark/benchmark-matmult.cpp:253:23: note: use function 'std::abs' instead
        float delta = abs(sum_of_Q4_result - sum_of_F32_reference);
                      ^~~
                      std::abs
1 warning generated.