cyl_bessel_j return NaN for low numbers

[agribov_no_s_p_a_m@…]

Calculate cyl_bessel_j<double, double> for v = 10.0 and x = 1.0e-100. Return value is NaN instead of 0.0. Problem happens in file: \math\special_functions\detail\bessel_jn.hpp Line: 92. ratio become NaN.

[John Maddock]

Confirmed, investigating.

[John Maddock]

(In [72221]) Update Bessel functions to use series representations when x is small compared to v and also to prevent the generation of NaN's in such cases. Update docs to match. Fixes #5560.

[John Maddock]

Reopened: there's still an issue for v < 0 and |v| >> x. Will be harder to fix though :-(

[John Maddock]

(In [72415]) Add new series evaluation forms for the Bessel J and Y functions. Also take extra care when evaluating to avoid spurious overflow/underflow. Changed check_series_iterations to know the real-number type so messages are properly formatted. Updated docs to match and regenerated. Fixes #5560.