Treetime

- calls to some random number generator errorred after the recent switch to treetime owned RNGs.
- default argument for clock-filter method had a spelling mistake
- a branch length optimization function didn't handle profiles==True correctly.

- rewrite of the vcf reading and writing code including many more tests, closer alignment with spec, and performance improvements by jameshadfield [PR 263](https://github.com/neherlab/treetime/pull/263)
- clock-filter command can now remove detected outliers with the additional flag `--prune-outliers`.

- fix division by zero error during GTR inference
- improve doc strings in parse dates
- tweaks to background shading in timetree plot function (`plot_vs_years`)
- allow to specify branches on which date confidence intervals are shown.

Previously, only a crude analysis of whether the divergence of tips roughly follows a linear trend was implemented. Tips that deviated too much from that regression line were flagged as outliers and this threshold was parameterized as number of interquartile distances of the distribution of residuals `n_iqd`.
This filter is not very sensitive and often misses misdated tips that severely distort the tree but still fall within the distribution of root-to-tip distances at that time.
To overcome this, we implemented a novel filtering method that fits a simple gaussian model of divergence accumulation.
Information on outliers is saved in a pandas DataFrame `self.outliers` of `TreeTime` and written to file as a tsv file when running treetime as command line tool.

Other fixes
* error when rate estimate is negative during the rate susceptibility calculation. Give hint in error message to specify the rate and its uncertainty explicitly.
* Fix bug [issue 250](https://github.com/neherlab/treetime/issues/250) introduced in 0.10.0 where treetime fails in absence of an alignment when trying to create an auspice json file. [PR #251](https://github.com/neherlab/treetime/pull/251)

* avoid probability loss at the end of the domain of distributions
* fix erroneous check for merger model.
* raise error when probability is lost.
* improve initial guess in branch length optimizations

0.10.0: add auspice.json output, drop python 3.6

* the output directory now contains a json file that is compatible with auspice.us. Both time scaled phylogenies and ancestral inferences can now be visualized and explored using auspice. Available colorings are "Date", "genotype", "Branch support", and "Excluded". See [PR 232](https://github.com/neherlab/treetime/pull/232) for details.
* move most function related to IO of the command line wrappers into a separate file.
* make TreeTime own its random number generator and add `--rng-seed` to control state in CLI. Any previous usage of `numpy.random.seed` will now be ignored in favor of `--rng-seed`. See [PR 234](https://github.com/neherlab/treetime/pull/234)
* add flag `--greedy-resolve` (currently default) as inverse to `--stochastic-resolve` with the aim of switching the default behavior in the future.
Add deprecation warning for `greedy-resolve`.
* tighten conditions that trigger approximation of narrow distribution as a delta function in convolution using FFT [PR 235](https://github.com/neherlab/treetime/pull/235).
* Drop support for python 3.6.
* Don't attempt to show figure when calling `Phylo.draw` to suppress matplotlib back-end warning.

* in cases when very large polytomies are resolved, the multiplication of the discretized message results in messages/distributions of length 1. This resulted in an error, since interpolation objects require at least two points. This is now caught and a small discrete grid created.
* increase recursion limit to 10000 by default. The recursion limit can now also be set via the environment variable `TREETIME_RECURSION_LIMIT`.
* removed unused imports, fixed typos
* add new way to resolve polytomies. the previous polytomy resolution greedily pulled out pairs of child-clades at a time and merged then into a single clade. This often results in atypical caterpillar like subtrees. This is undesirable since it (i) is very atypical, (ii) causes numerical issues due to repeated convolutions, and (iii) triggers recursion errors during newick export. The new optional way of resolving replaces a multi-furcation by a randomly generated coalescent tree that backwards in time mutates (all mutations are singletons and need to 'go' before coalescence), and merges lineages. Lineages that remain when time reaches the time of the parent remain as children of the parent. This new way of resolving is much faster for large polytomies. This experimental feature can be used via the flag `--stochastic-resolve`. Note that the outcome of this stochastic resolution is stochastic!