Fix the calculation for the updated belief

[rikhuijzer]

The order was reversed as could be seen by computing the posterior distribution for one heads or one tails.
For example, if you calculated the posterior for a sample of size one and containing one heads, the posterior plot would show that the parameter is most likely to be 0 whereas the plot shows that an higher value means a higher probability for heads.

[cpfiffer]

This won't fix anything, unfortauntely. Currently the site is not served from this repo, but from TuringLang/Turing.jl.

Tutorials come upstream and any fixes should correct the Jupyter notebooks here: https://github.com/TuringLang/TuringTutorials

[cpfiffer]

Also this fix doesn't seem to need to be relevant for the main site, since the change is already present there: https://turing.ml/dev/tutorials/0-introduction/

[rikhuijzer]

Also this fix doesn't seem to need to be relevant for the main site, since the change is already present there: https://turing.ml/dev/tutorials/0-introduction/

The output on that link has

updated_belief = Beta(prior_belief.α + heads, prior_belief.β + tails)

whereas I propose

updated_belief = Beta(prior_belief.α + tails, prior_belief.β + heads)

I have now proposed this change again at TuringLang/docs#91.