Pycomex

-------------------

- Fixed an issue with the experiment command line argument parsing where parameters of sub experiments (defined by
Experiment.extend) would not show up in the help message. Solved by re-constructing the ExperimentArgumentParser
in the extend method.
- The help command will now also show the default values of experiment parameters (if their string representation is
below a certain length.)
- Cleaned up the unittests. In this version all of the unitests actually passed before the publish
- "notify" plugin
- can now actually be disabled using the ``__NOTIFY__`` special parameter
- "weights_biases" plugin
- Now actually correctly handles when the ``WANDB_PROJECT`` parameter is incorrectly set.

-------------------

- Ported the notification implementation as a plugin instead of having it in the main code
- Clicking a notification will now open the experiment record folder in the file explorer
- Disabled the notifications for the windows.

-------------------

- Removed the Python 3.10 Union type hint from the experiments module to restore compatibility with Python 3.9
- Added a try block for the loading of plugins such that they re-cast errors as warnings and don't actively break
the system such as when an import error in a plugin module occurs (aka we dont want to have to include all the plugin
dependencies in the main package)

-------------------

- In the wandb plugin: Moved the login of the project into the "after_initialize" hook because there was an issue
with not being able to overwrite the WANDB_PROJECT parameter for sub experiments...

-------------------

- Extended the ``Experiment.track`` method to be able to track figures as well by storing them into a specific
folder within the experiment archive folder.
- The ``Experiment.metadata`` dict now contains the "__track__" entry which is used to store the names of all
the tracked quantities.
- Added the ``plot_track`` plugin which is mainly executed after each experiment and will automatically plot all
the tracked quantities into separate visualizations. Numeric quantities will be plotted as line plots and
figures will be stitched together into a video.
- Added ``moviepy`` to the dependencies

-------------------

- Added the ``Experiment.track_many`` method which can be conveniently used to track multiple artifacts at once
- Changed the track function generally to store the values in a list instead of replacing the value each time.