Cobra

Other

* Update libSBML dependency to `python-libsbml==5.19.0`.
* Docstring and style changes.

New features

* There is a new function inspired by
[cameo](https://github.com/biosustain/cameo/) that can load models directly
from [BiGG](http://bigg.ucsd.edu/) or
[BioModels](https://www.ebi.ac.uk/biomodels/) and by default will cache them
locally (995), for example,

py
from cobra.io import load_model

ecoli = load_model("iML1515")
ylipo = load_model("MODEL1604280017")


Fixes

* Over a series of pull requests, synchon has been modernizing parts of the
codebase for Python 3.6+ (993, 999, 1000, 1003)
* Upgrade the dependency python-libsbml-experimental to 5.18.3

We now have an official code of conduct and a committee of three people to handle
reports! You can contact them at [conductopencobra.org](mailto:conductopencobra.org).
You can also contact them individually at:

* [Maureen Carey](mailto:MAC9JCvirginia.edu)
* [Svetlana Kutuzova](mailto:svegalbiosustain.dtu.dk)
* [Greg Medlock](mailto:GLM5UHvirginia.edu)

Many thanks to them for taking on this responsibility.

New features

- Deletion result DataFrames as returned by functions from `cobra.deletions`
now have a new `knockout` accessor. See the docs on usage examples.
- All summaries are now persistent objects. The same summary can be displayed in
multiple ways. The underlying data for the summary can also be accessed.
- The model summary now displays elementary (by default carbon) uptake and secretion.
- The JSON schema is now available as a stand-alone JSON file.

Fixes

- Remove the frozenset indexing in deletion DataFrames that is now unsupported
in pandas.
- Summaries for models, metabolites, and reactions are now calculated instantaneously
and thus also work with model contexts.
- JSON models can now contain infinity and NaN values as bounds.
- The Matlab bridge for reading Matlab models should be restored. Any testing or
problem reports are greatly appreciated.

Deprecated features

- We are progressively updating our codebase. That means that more and more parts will
require Python 3.6+.

Backwards incompatible changes

- Deletion result DataFrames have no frozenset index anymore but now carry the
deleted elements in the `ids` column.