Generalize & improve `test_propodds`

[tsostarics]

Test propodds works fine for most of my use cases but a few small changes would iron out a few issues:

• Allow for a dependent variable with name other than scale_response
• Add argument for link function to use in the glm call (currently defaults to logit, but it needs the probit option too)
• add an option to include a CLM or CLMM object. The goal with this would be to compare the CLMM estimates with their associated confidence interval to add to the plot. Currently it defaults to the mean of the estimates from the individually fit logistic regressions, which can be the default behavior if no CLMM is provided. This may require some (tbh much needed) refactoring to modularize this process.
• investigate corrections for response bias (may not be possible, requires looking through literature extensively. May be above my pay grade. If someone out there is reading this help on this would be appreciated.)

[tsostarics]

I forgot about this issue, but the first three points here are already implemented:

• The full formula is passed to test_propodds
• A link argument is included, defaulting to logit
• An ordinal model argument is included, should use the same formula passed to test_propodds

The final point requires additional specification of the discrimination parameter, which is easier to do using Bayesian models in brms.