setting prior fails if there are no fixed parameters in the model

[venpopov]

After #152, a new bug appears if we try to predict all parameters of a model and none are left fixed:

formula <- bmf(mu ~ 1,
               c ~ 1,                
               kappa ~ 1)   
default_prior(formula, OberauerLin_2017, sdmSimple('dev_rad'))

Error: Processing arguments of 'set_prior' has failed:
Error in (function (..., row.names = NULL, check.rows = FALSE, check.names = TRUE,  : 
  arguments imply differing number of rows: 0, 1

If at least one parameter is fixed to a constant it works fine:

formula <- bmf(mu ~ 1,
               c ~ 1,                
               kappa = 2)   
default_prior(formula, OberauerLin_2017, sdmSimple('dev_rad'))

This is caused by the "fixed_pars_priors", which throws an error if there are no fixed parameters. We should add a condition to execute this only if fixed parameters exist

[venpopov]

I have already fixed this on the bugfix branch, but I noticed that we don't have a default prior for mu, and that if users predict mu, it gets the default brms prior, which turns out it's quite strange for the von_mises distributions - paul-buerkner/brms#1631

Before adding a pull request, I will see if I can add default priors for the mu parameters in our models. I suspect this might require some adjustments to ensure that they do not overwrite the constant priors when they are fixed by default.