Function tgamma_impl 

pub fn tgamma_impl<P, E, V, const N: usize>(
    z_in: V,
    l: &Lanczos<E, N>,
    int_cap: E,
    ln_max: E,
) -> V
where
    P: Policy,
    E: FloatElementWithBits,
    V: FloatVectorWithBits<Element = E> + SpecializedSpecialMath<E>,

Shared tgamma implementation for all real element types.

Covers only the precision >= Average path; the low-precision shortcut through lgamma_r stays at the call site, because on f32 lgamma_r has its own low-precision Pade branch that this module deliberately does not know about.

• int_cap: the largest integer whose factorial is finite in E (36 for f32, 172 for f64). Bounds the integer fast-path loop.
• ln_max: ln(E::MAX), the overflow threshold for the pow in the main term.