Wknml

New Features:
- added docstrings and auto-generated documentation. (see `/docs`)
- further internal changes:
- added a `pyproject.toml` configuration file
- added `poetry` for dependency management and publishing
- added Python formatting (`black`) and documentation checking to CI
- updated README to include more examples

Breaking Changes:
- added requirement for Python 3.6+
- refactored all `NML`, `Node`, `Tree`, `Comment`, ... objects to inherit directly from `namedtuple`

* Auto-versioning and auto-publishing

- Added `nml_generation` model to generate NMLs from networkX graphs
- Fixed that offsets are also persisted
- Made node radius optional

New Features:
- wknml now uses a streaming NML reader and writer, so large NML files should no longer cause memory problems.
- Optional property support: Optional properties may be omitted and will no longer be persisted when dumping the NML, which allows to write smaller, more compact NML files. For example, only the `id`, `radius`, and `position` of a `Node` are mandatory, the remaining 6 properties are optional and can be omitted.

Breaking Changes:
- `parse_nml(file)` now takes a file instance as its parameter.
- `dump_nml` has been renamed to `write_nml(file, nml)` which now takes a file instance as the first parameter and an NML instance as the second. The method will take care of writing the dump of `nml` to the `file`.
- A `Group` now needs a third parameter which contains its children and is of type `List[Group]`, so `[]` if the group has no children.