Add more options to control the length of normals in MeshNormals visual

[asnt]

Use cases:

1. To get normals that are about the length across a triangle of the mesh:

MeshNormals(..., length_method='median_edge', length_scale=1.0)
MeshNormals(...)  # equivalent to the previous line (default values)

2. To get normals that are about 10% the size of the mesh:

MeshNormals(..., length_method='max_extent', length_scale=0.1)

[djhoese]

Would it be possible for you to add a test?

Also, what about some examples like the ones you listed in this PR's description but in an Examples section in the docstring. Like:

    **kwargs : dict, optional
                Extra arguments to define the appearance of lines. Refer to
                :class:`~vispy.visuals.line.LineVisual`.

    Examples
    -------------
    Normals at about the length of a triangle:

    ```
    MeshNormals(..., length_method='median_edge', length_scale=1.0)
    ```

    For normals at about 10% the size of the mesh:

And so on. Thoughts?

[asnt]

Would it be possible for you to add a test?

I will look at that next.

Also, what about some examples like the ones you listed in this PR's description but in an Examples section in the docstring.

Good idea. I had a try.

And so on. Thoughts?

I was thinking about making it more convenient to use this with a MeshVisual. Currently

mesh = MeshVisual(...)
normals = MeshNormals(mesh.mesh_data, ...)

Two ideas:

1. normals = MeshNormals(visual=mesh)
2. mesh = MeshVisual(..., normals='vertex')

With option 1, it is obvious where to configure the normals, at construction time or through normals.
With option 2, it is not obvious where to go to configure the normals and we'd need to expose mesh.normals, for example.

[asnt]

@djhoese In the tests, I set up a camera instead of defining the scene transformations manually. It seems simpler to maintain in my opinion. Is there maybe another reason for the manual transformations instead?

[asnt]

The error is when initializing the vispy-tests environment with micromamba:

/usr/bin/bash -c source /home/runner/.bash_profile && micromamba list
ERROR   Cannot activate, prefix does not exist at: /home/runner/micromamba/envs/vispy-tests
Error: The process '/usr/bin/bash' failed with exit code 1

I don't see at first how this could be linked to the last commit.

[kmuehlbauer]

I don't see at first how this could be linked to the last commit.

@asnt Looks like a spurious thing. Maybe just rerun it will help?

[asnt]

@asnt Looks like a spurious thing. Maybe just rerun it will help?

Thanks @kmuehlbauer . I don't think I can relaunch the tests. Let's see with the new commit.

[asnt]

@djhoese The tests now cover all parameters. For length_method, the check is for it not to raise. Not sure how better to test it.
Any further suggestion?

[djhoese]

This looks really good. Nice job on the tests. One request: could you have the first Example in your docstring show you creating the Mesh visual and passing the meshdata instance to the normals visual?