Function compute_gae_batched_f32 

pub fn compute_gae_batched_f32(
    rewards: &[f32],
    values: &[f32],
    dones: &[f32],
    last_values: &[f32],
    n_steps: usize,
    gamma: f32,
    gae_lambda: f32,
) -> (Vec<f32>, Vec<f32>)

Batched GAE in f32 — avoids f64 conversion overhead from Python.

Same layout as compute_gae_batched but operates on f32.

Panics

Panics in debug builds if input slice lengths do not match n_envs * n_steps.