Collection of fixes and enhancements for branch `separate_scaling`

[fweber144]

This is a more or less loose collection of fixes and enhancements on top of branch separate_scaling that I found while working with priorsense, especially in a predictive setting.

The commit messages should provide details where necessary. If you need more information, just let me know.

I have also added myself as a contributor. If you don't want this, simply revert the corresponding commit.

I have not run R CMD check or the unit tests at all. Please tell me if I should do so.

[n-kall]

Thanks @fweber144 for all of this! I'll take a closer look but it's likely good to merge as is

[fweber144]

For the prediction part in powerscale_sequence.priorsense_data(), I'm now realizing that my changes might be unnecessarily complicated. The approach in powerscale.priorsense_data() and powerscale_gradients.priorsense_data(), namely to initially assign only base_draws <- x$draws, then appending the predictions, and then to call posterior::subset_draws() would probably be more straightforward and would avoid the extra definition of variable_base. But I'm not sure if that approach is really transferable to powerscale_sequence.priorsense_data().

[n-kall]

These changes look all good to me! I'll merge now and if we want to change the prediction part as you said, we can do that with another PR

[fweber144]

Thanks!