Does not try to retain assets, or defined asset values, and loses the types of non string customizations currently.

- Add missing section to md result we cannot roundtrip right now
- Move imports to a dedicated sections with a comment
- Using `ast.literal_eval`, customizations that are literals are parsed back from strings to their native type

now attempts to list assets found in the 4 key folders from prairielearn

Need to flatten the namedtuple to a list

- Now copies asset files
- Allows for an output directory and filename to be passed
- Validates input paths are directories and not files
- Improves argument checking

- Need to prevent `"..."` from being evaled to the `Ellipses` literal
- Remove the `__pycache__` directory created by importing and executing the `server.py` file
- Fix `server.py` parsing
- Fix asset copying
- Add `pl-rich-text-editor` to list of supported pl inputs

Also minor updates to overall test layout to make round trip testing work nicer

Now should upload md generated from pl files on failure

…metadata field of the comment object in `info.json`

Start fixing bugs/issues in tests causing failures

…if one on a question already fails.

…rkdown