fix divergence_exact's diagonal sum

[ventusff]

In run_nerf_helpers.py: L71 divergence_loss, the original code to calculate the diagonal of jacobian matrix is wrong.

I'm guessing since it's using divergence_approx function by default, this error does not affect the default training behavior.
But still it will affect the behavior when setting exact=true for training.

example:

jac[0,...]=
        [[  1.5425,   8.1350,   5.7477],
        [ -1.1500, -14.2937,  -9.5665],
        [ -0.5321,  11.8239,   5.3827]]

jac.view(jac.shape[0], -1)[0,...]=
         [  1.5425, 8.1350, 5.7477, -1.1500, -14.2937, -9.5665, -0.5321, 11.8239, 5.3827]

# wrong way
jac.view(jac.shape[0], -1)[0, :: jac.shape[1]]=
         [ 1.5425, -1.1500, -0.5321]

# correct way
jac.view(jac.shape[0], -1)[0, :: (jac.shape[1]+1)]=
         [  1.5425, -14.2937,   5.3827]

[facebook-github-bot]

Hi @ventusff!

Thank you for your pull request and welcome to our community. We require contributors to sign our Contributor License Agreement, and we don't seem to have you on file.

In order for us to review and merge your code, please sign at https://code.facebook.com/cla. If you are contributing on behalf of someone else (eg your employer), the individual CLA may not be sufficient and your employer may need to sign the corporate CLA.

If you have received this in error or have any questions, please contact us at cla@fb.com. Thanks!

[ventusff]

OK, already signed now

[ventusff]

re-open with CLA signed