Changing the prior for negative binomial observations

[SamuelBrand1]

This addresses #29 using the link given by @seabbs in that issue.

Main change is here

Rt-without-renewal/EpiAware/src/observation-models.jl

Lines 46 to 61 in a048da5

	#Parameters
	neg_bin_cluster_factor ~ observation_model.neg_bin_cluster_factor_prior
	#Predictive distribution
	expected_obs = observation_model.delay_kernel * I_t .+ pos_shift
	if ismissing(y_t)
	y_t = Vector{Int}(undef, length(expected_obs))
	end
	for i in eachindex(y_t)
	y_t[i] ~ NegativeBinomialMeanClust(
	expected_obs[i], neg_bin_cluster_factor^2
	)
	end

The point is that we want to have a half-normal family prior for a standard deviation. In this case, its the standard deviation of the latent Gamma process implicit in the Negative Binomial definition. All this is based on this classic paper.

I've update the getting started note and the sampler seems happy with this new prior.

Closes #29