Features
* ancestral: Adds a new flag `--output-sequences` and logic to support saving
ancestral sequences and leaves from the given tree to a FASTA file. Also adds a
redundant, more specific flag `--output-node-data` that will replace the current
`--output` flag in the next major version release of augur. For now, we issue a
deprecation warning when the `--output` flag is used. Note that FASTA output is
only allowed for FASTA inputs and not for VCFs. We don't allow FASTA output for
VCFs anywhere else and, if we did here, the output files would be very large.
[See PR 293](https://github.com/nextstrain/augur/pull/293)
* frequencies: Allow `--method kde` flag to compute frequencies via KDE kernels.
This complements existing method of `--method diffusion`. Generally, KDE
frequencies should be more robust and faster to run, but will not project as
well when forecasting frequencies into the future.
[See PR 271](https://github.com/nextstrain/augur/pull/271)
Bug fixes
* ancestral, traits, translate: Print warning if supplied tree is missing internal
node names (normally provided by running `augur refine`).
[See PR 283](https://github.com/nextstrain/augur/pull/283)
* Include pip in Conda enviroment file.
[See PR 309](https://github.com/nextstrain/augur/pull/309)
Documentation
* Document environment variables respected by Augur
Development
* Remove matplotlib and seaborn from `setup.py` install. These are still called a
few places in augur (like `titers.validate()`), but it was deemed rare enough
that remove this from `setup.py` would ease general install for most users.
Additionally, the ipdb debugger has been moved to dev dependencies.
[See PR 291](https://github.com/nextstrain/augur/pull/291)
* Refactor logic to read trees from multiple formats into a function. Adds a new
function `read_tree` to the `utils` module that tries to safely handle reading
trees in multiple input formats.
[See PR 310](https://github.com/nextstrain/augur/pull/310)