Bayes-ab

This release adds several features to the DiscreteDataTest class, including:

- The ability to plot the posterior distributions for each state for each variant
- Confidence intervals for both the mean and the relative probability of each state
- Uplift-vs.-control
- Improved readability of test output

This release also introduces the sample mean into the output for binary, Poisson, and normal tests, to allow simpler comparison to the Bayesian posterior.

Finally, this release corrects critical errors in v0.4.2, where the 95% HDI for both the Poisson test and the binary was actually calculated as the 90% HDI.

This release adds several features to the `NormalDataTest` class, including:

- The ability to plot the joint prior distribution for $\mu$ and $\sigma^2$
- The ability to plot posterior and difference distributions
- Confidence intervals for both $\mu$ and $\sigma$
- Uplift-vs.-control and precision metrics
- Improved readability of test output
- Exposed sample data for samples from the gamma distribution, for $\frac{1}{\sigma^2}$

This release also corrects two critical errors:

- The first was present in v0.3.0, in the calculations for the posterior mean for normal data
- The second was present in v0.4.0 and v0.4.1, in the plotting—_but not calculation_—of the prior distributions

This small release fixes several broken links across various documentation files.

This release rebrands the the `bayesian-testing` package as `bayes_ab` and incorporates a variety of improvements:
- Poisson testing
- The option to check metric estimates against closed-form solutions (Poisson and Bernoulli tests only)
- The ability to plot posterior and difference distributions (Poisson and Bernoulli tests only)
- The option to assess whether to continue or terminate the test via a standard decision rule (Poisson and Bernoulli tests only)
- Confidence intervals and uplift-vs.-control metrics (Poisson and Bernoulli tests only)
- Improved readability of test output (Poisson and Bernoulli tests only)
- Exposed sample data

This release also corrects a critical error present in v0.3.0, in the calculation of the Bernoulli positive rate.