change redshift interpolation to log(1+z)

[brittonsmith]

This is a reissue of PR #105, which I've taken over for BK. I've merged with the latest changes and made just a couple modifications, but I will leave the rest open for discussion.

The gist of this PR is that it changes the interpolation over redshift for the UV background and tabulated cooling to interpolation over log(1 + z). This is an improvement since all of our redshift-dependent tabular data is evenly spaced in log(1 + z). My recollection is that it was decided at a user/dev meeting to go with this without maintaining an option for the old behavior as it is a general improvement.

This PR also includes changes to the tiny8 constant and setting edot to zero when it's very small. These I am less sure of these, but will leave them open for discussion.

[brittonsmith]

This fails the cooling_rate.py tests with average relative differences of just over ~1e-7 in some of the species densities. I think this is acceptable and can update the test answers when the rest is ready to go.

[gregbryan]

Looks good to me although before merging, it would be good to get another pair of eyes on the questions I raised in my review. Great job!