Chromatic noise models cleanup

[blarsen10]

A few changes which allow enterprise_extensions to be used to run all of the custom noise models from the recent NANOGrav submission:

• Allows annual chromatic signals to be modeled as a GP instead of a deterministic signal
• Fixes a bug where the free chromatic noise component would not vary the chromatic index when using a squared exponential kernel
• Fixes a few syntax errors in the dm_noise_block
• Allows more options for the chromatic annual signal to be set in the model_singlepsr_noise function
• Cleans up the exponential dip naming scheme in model_singlepsr_noise, which was not very flexible before