User-friendly errror messages for prob macro

[awellis]

I tried to obtain the point-wise likelihood for a model containing a predictor variable:

@model AnovaModel(y, group) = begin
    n_groups = length(unique(group))

    σ ~ truncated(Cauchy(0, 3), 0, Inf)
    α ~ filldist(Normal(0, 10), n_groups)
    μ = α[group]
    @. y ~ Normal(μ, σ)
end

chain = sample(AnovaModel(y, group), NUTS(), 2000, save_state = true)

but I forgot to include the predictor on the right-hand side:

logprob"y = y | chain = chain"

instead of

logprob"y = y | group = group, chain = chain"

The following error message is not very helpful: ERROR: AssertionError: all(valid_arg, getargnames(modelgen)).
Maybe the user could be encouraged to add the missing predictors, or could they be conditioned on by default?

[yebai]

Closed in favour of #356