To get you started, you should be able to use some straight forward Vector3 calculations to find a center point between all of the players' positions.

From there you should make that Vector3 the target of your camera so it follows/centers on it. To tackle the zoom, since you should know all the players positions already, you could find the 2 who are farthest apart from each other and use that when calculating where your camera should be on the Y axis.

[EDIT]

In response to your comment...

You could use the Mesh class to create a mesh that uses all of the player's transform.position Vector3's as it's vertices.

From there, you can use the Mesh.bounds property to return an Axis Aligned Bounding Box (AABB) which you can use the Bounds.center property (a Vector3) as the focal point of your camera. Then, because you have an AABB, the larger the number, the further out your camera should be, and vice-versa.

Here's a little something to get you started:

using UnityEngine;
using System.Collections;

public class FramePlayers : MonoBehaviour {

    //set this in the inspector
    public Transform[] playerLocs; 

    private Mesh m;

    void Update () {

        //need a new mesh every frame
        m = new Mesh();

        //new set of vector3s aswell
        Vector3[] v3s = new Vector3[playerLocs.Length];

        //populate
        for(int i = 0; i < playerLocs.Length; i++)
        {
            v3s[i] = playerLocs[i].position;
        }

        //asign them to the mesh
        m.vertices = v3s;

        //print the resulting AABB's center. I tested this with an even and odd number of transforms.
        print(m.bounds.center + "is the center of this mesh.");

    }
}

Hope that helps.

==