v4.5.0 (21 November 2022)

• Added the generalised hypergeometric function. (#2510)
• Added var overloads for pow(), owens_t(), log_inv_logit . (#2546, #2787, #2806)
• Add tuple overload and cleanup code for promote_scalar. (#2706)
• New gradient function allows evaluated gradient to be stored into an array, enabling memory reuse when stan/math called via a FFI. (#2743)
• Added function for casting double and std::vector<double> types to int and std::vector<int> types(#2771)
• Added hypergeometric_2F1 function. (#2792)
• Fix bug with vectorised pow() incorrectly forcing Eigen::Matrix return instead of Eigen::Array. (#2793)
• Exposed the hypergeometric_3F2 function and improved its numerical stability. (#2797)
• Fixed incorrect return from log_inv_logit_diff with positive infinity first argument. (#2798)
• Added reverse-mode specializations for fft2 and inv_fft2. (#2800)
• Removed the cause of the out-of-order initializer compiler warning.(#2808)
• Vectorized atan2().(#2812)
• Added complex_schur_decompose_t and complex_schur_decompose_u.(#2814)
• Vectroized the conj() function. (#2817)
• Stan's algebraic solvers now support variadic arguments. (#2820)
• Fixed forward mode autodiff for FFT functions. (#2821)
• Added log_sum_exp_signed function for computing log_sum_exp while respecting signs of arguments and tracking the sign of the result. (#2829)
• stan_print can now print std::tuple types. (#2835)
• stan::math::minus can now accept std::vector. (#2840)