Function dispatch_tri_solve 

pub fn dispatch_tri_solve(
    encoder: &mut CommandEncoder,
    registry: &mut KernelRegistry,
    device: &DeviceRef,
    l: &MlxBuffer,
    b: &MlxBuffer,
    x: &MlxBuffer,
    params_buf: &MlxBuffer,
    p: TriSolveParams,
) -> Result<()>

Dispatch a lower-triangular unit-diagonal solve.