Tsdate

**Features**

- The algorithm now operates completely in unscaled time (in units of generations) under
the hood, which means that `tsdate.build_prior_grid` now requires the parameter
`Ne`.
- Users now have access to the marginal posterior distributions on node age by running
`tsdate.get_dates`, though this is undocumented for now.

**Bugfixes**

- A fix to the way likelihoods are added should eliminate numerical errors that are
sometimes encountered when dating very large tree sequences.

**Features**

- Two new methods, `tsdate.sites_time_from_ts` and `tsdate.add_sampledata_times`,
support inference of tree sequences from non-contemporaneous samples.
- New tutorial on inferring tree sequences from modern and historic/ancient samples
explains how to use these functions in conjunction with `tsinfer`.
- `tsdate.preprocess_ts` supports dating inferred tree sequences which include large,
uninformative stretches (i.e. centromeres and telomeres). Simply run this function
on the tree sequence before dating it.
- `ignore_outside` is a new parameter in the outside pass which tells `tsdate` to
ignore edges from oldest root (these edges are often of low quality in `tsinfer`
inferred tree sequences)
- Development environment is now equivalent to other `tskit-dev` projects

- Improve user experience with more progress bars and logging.
- Slightly change traversal method in outside and outside maximization algorithms,
this should only affect inference on inferred tree sequences with large numbers
of nodes at the same frequency.
- Improve reporting of current project version
- Use appdirs for default caching location
- Prevent dating tree sequences with dangling nodes

- Bugfix release: resolve issue with precalculating prior values.

Early Alpha release made available via PyPI for community testing and evaluation.

Please don't use this version in published works.