Maybe small improvement for accuracy for cos for regular angles

@seiko2plus - Came here out of curiosity from https://github.com/numpy/numpy/pull/29699#issuecomment-3630501457, to see how things get done. For cos, the following transformation is done before any real calculation:
https://github.com/numpy/numpy-simd-routines/blob/1acd14571b9f58072d1f7f17886945000d78b56a/npsr/trig/high-inl.h#L35-L43

This would seem to loose some precision for the common case where angle input will be between -pi and pi, by moving those angles to pi/2 to 3*pi/2. Instead, one can move those angles to near zero and thus get optimal precision near both crossings by using that cos(x) = sin(pi/2 - |x|) , i.e. do something like,

  // Transform cosine to sine using identity cos(x) = sin(π/2 - |x|)
  // where |x| ensures optimal precision near both -pi/2 and pi/2.
  const V half_pi = Set(d, data::kHalfPi<T>);
  V x_trans = x;
  if constexpr (OP == Operation::kCos) {
    x_trans = And(abs_mask, x);
    x_trans = Sub(half_pi, x_trans);
  }
  const V x_abs = And(abs_mask, x_trans);
  const V x_sign = AndNot(x_abs, x_trans);

Start in npsr/trig/high-inl.h at lines 35–43 and trace the cosine preprocessing path in the SIMD routine. Compare the existing transformation with the proposed cos(x) = sin(π/2 - |x|) approach, focusing on inputs between -pi and pi. Done means the cosine path uses the accuracy-preserving transformation without regressing the surrounding sine/cosine handling.