Phenopype

-----
bugfix:
- removed "zoom memory" when running Pype, as this wasn't working properly
- Project.check_files() now also accounts for missing filenames in the project attributes file
- "show_nodes=true" argument for all polygon-related measurement tools (polygon mask, rectangle mask, polyline)

-----
- new function: class method project.check_files; will all project files for completeness by comparing the images in the
"data" folder to the file list the project attributes. Will attempt to
fix discrepancies, but ask for feedback first.
- project class method add_files now takes n_max argument, to limit the amount of images being added to the project

-----
bugfix:
- skip=True in pype now works with any produced annotation or canvas file (or specified annotation, e.g. skip="contours")
- polyline would be converted to polygon
- internal feedback argument would be accidentally included in pype config

-----
bugfix:
- plugin imports not working
- added manifest to include assets folder with html color overview

-----
bugfix:
- added "gray" option back to select_canvas
- images added to the project with "copy" or "mod" mode will, by default, keep their file ending. there is also a new resizing option for mod, which will downsize the images to the specified "resize_max_dim" while maintaining aspect ratio
new functionality:
- plugins! phenopype can now check for external packages and, if available, load functions that utilize them. currently, this includes "phenomorph" for automatic landmark detection (using a dlib backend), and keras (using a tensorflow backend) for deep learning based segmentation. if the packages are not installed to the same python environment as phenopype they simply wont work (with a warning).
- option "visualize" for the Pype class is now "autoshow" and can be used independently from "feedback"
- new function: pp.print_colours() will show the names from W3 colour palette that can be used for all functions accepting colour arguments as strings

-----
bugfix:
- training data exported for "ml-morph" framework wasn't formatted correctly
- new training data export for convolutional neural networks in keras (using "keras-cnn-semantic")