MultyBody with FixedJoint causes a panic when only one part is attached to it

The following program panics with the message below:

```rust
let mut world = World::new();

let shape = ShapeHandle::new(Cuboid::new(Vector3::repeat(0.5)));
let collider = ColliderDesc::new(shape)
.density(1.0);

let top_joint = FixedJoint::new(Isometry3::identity());
let mut body_desc = MultibodyDesc::new(top_joint)
.collider(&collider)
.parent_shift(Vector3::repeat(1.0));

/*let sub_joint = PrismaticJoint::new(Vector3::x_axis(), 0.0);
body_desc.add_child(sub_joint)
.set_parent_shift(Vector3::repeat(1.0))
.add_collider(&collider);*/

body_desc.build(&mut world);

Testbed::new(world).run()
```

```
thread 'main' panicked at 'Matrix slicing out of bounds.', /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/nalgebra-0.18.0/src/base/matrix_slice.rs:227:9
stack backtrace:
0: std::sys::unix::backtrace::tracing::imp::unwind_backtrace
1: std::sys_common::backtrace::print
2: std::panicking::default_hook::{{closure}}
3: std::panicking::default_hook
4: std::panicking::rust_panic_with_hook
5: std::panicking::begin_panic
at /usr/src/rustc-1.34.2/src/libstd/panicking.rs:412
6: nalgebra::base::matrix_slice::>::assert_slice_index
at ./<::std::macros::panic macros>:3
7: nalgebra::base::matrix_slice::>::columns_generic
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/nalgebra-0.18.0/src/base/matrix_slice.rs:410
8: nalgebra::base::matrix_slice::>::fixed_columns
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/nalgebra-0.18.0/src/base/matrix_slice.rs:391
9: nalgebra::base::matrix_slice::>::column
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/nalgebra-0.18.0/src/base/matrix_slice.rs:360
10: nalgebra::base::blas::>::quadform_with_workspace
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/nalgebra-0.18.0/src/base/blas.rs:1487
11: nalgebra::base::blas::>::quadform
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/nalgebra-0.18.0/src/base/blas.rs:1532
12: >::update_inertias
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/nphysics3d-0.11.1/src/object/multibody.rs:465
13: >::update_dynamics
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/nphysics3d-0.11.1/src/object/multibody.rs:381
14: as nphysics3d::object::body::Body>::update_dynamics
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/nphysics3d-0.11.1/src/object/multibody.rs:841
15: >::step
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/nphysics3d-0.11.1/src/world/world.rs:238
16: ::step
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/nphysics_testbed3d-0.5.0/src/testbed.rs:552
17: kiss3d::window::window::Window::do_render_with_state
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/kiss3d-0.19.0/src/window/window.rs:597
18: kiss3d::window::window::Window::render_loop::{{closure}}
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/kiss3d-0.19.0/src/window/window.rs:581
19: ::render_loop
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/kiss3d-0.19.0/src/window/gl_canvas.rs:67
20: kiss3d::window::canvas::Canvas::render_loop
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/kiss3d-0.19.0/src/window/canvas.rs:31
21: kiss3d::window::window::Window::render_loop
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/kiss3d-0.19.0/src/window/window.rs:581
22: nphysics_testbed3d::testbed::Testbed::run
at /home/sven/.cargo/registry/src/github.com-1ecc6299db9ec823/nphysics_testbed3d-0.5.0/src/testbed.rs:243
23: reproduce::main
at src/reproduce.rs:27
24: std::rt::lang_start::{{closure}}
at /usr/src/rustc-1.34.2/src/libstd/rt.rs:64
25: std::panicking::try::do_call
26: __rust_maybe_catch_panic
27: std::rt::lang_start_internal
28: std::rt::lang_start
at /usr/src/rustc-1.34.2/src/libstd/rt.rs:64
29: main
30: __libc_start_main
31: _start
```

When I comment in the part that adds the sub part on a `PrismaticJoint` the panic goes away. I guess having a `FixedJoint` with only one part below it isn't really useful, as it could also be achieved by mounting a `RigidBody` to the ground, but I think at least the panic should give a proper message about the problem, as this is quite hard to debug.