Make `general_mat_mul`/`general_mat_vec_mul` more discoverable

Start with the linked general_mat_mul and general_mat_vec_mul documentation in the linalg module, then read the docs for the dot methods and Dot trait. Clarify whether the change should only improve documentation or also add the proposed GeneralDot trait and ArrayBase methods; done means the agreed functionality is discoverable and documented.

The general_mat_mul and general_mat_vec_mul functions are not easy to find, since they're in the linalg module while most other functionality is available either at the top level of the crate or as methods on the array types.

At a minimum, we should mention them in the docs for the dot methods and Dot trait.

However, I think it would also be useful to expose the functionality as methods on ArrayBase. I'd suggest adding a trait like this:

/// General matrix multiplication.
pub trait GeneralDot<A, Rhs1, Rhs2> {
    /// Perform general matrix-matrix multiplication, modifying `self` in-place.
    ///
    /// Compute C ← α A B + β C, where `self` is C.
    ///
    /// The array shapes must agree in the way that if `a` is *M* × *N*, then
    /// `b` is *N* × *K* and `self` is *M* × *K*.
    ///
    /// ***Panics*** if array shapes are not compatible<br>
    /// *Note:* If enabled, uses blas `gemm` for elements of `f32, f64` when memory
    /// layout allows.  The default matrixmultiply backend is otherwise used for
    /// `f32, f64` for all memory layouts.
    fn mul_add_scaled_dot_inplace(
        &mut self,
        beta: A,
        alpha: A,
        a: &Rhs1,
        b: &Rhs2,
    );
}

and then add mul_add_scaled_dot_inplace methods which use this trait to ArrayBase<S, Ix1> and ArrayBase<S, Ix2>. What do you think?