This is the beginning of a PR to fix #13

new

• __main__.py to make the kafl_fuzzer package executable (python -m kafl_fuzzer)
• scripts moved into the kafl_fuzzer package, refactoring later on
• subcommands added in config.py
• using the set_defaults to automatically associate and call each subcommand's entrypoint.

in cmdline.py:

• add a way to associate command line args <-> config file options by replacing confuse with dynaconf
  • the previous solution required to peer into the private attributes of the parser, which is not reliable long term
  • the new solution proposes to delegate the parameters validation from argparse to Dynaconf
  • the validators are placed in settings.py and associated with the dynaconf settings.
    
  • the configuration file precedence is defined by the app_settings_files() function in settings.py, and already integrates loading the KAFL_CONFIG_FILE environment variable
• the new unified main look like this:
  

So the new flow looks like this:

1. dynaconf is loading the settings files, according to the load order defined in settings.py

2. parser.parse_args() parses the command line and returns a Namespace object

3. this object is passed to update_from_namespac() function, defined in settings.py
   

4. this function will update the dynaconf settings by extracting a dict from the Namespace object and updating the settings dict.

5. the validation step comes in, and the validate function is called. Due to bug we had to write our own validation function and manually call the Validator cast functions.
   

6. the subcommand entrypoint is finally called in __main__.py

TODO

• add documentation
• add --help
• check subcommands standard args ( kafl {mcat,gui,plot} do not take the standard arguments. mcat takes a file, gui and plot take a workdir (as argument, not -w option) )
• discuss how the command line should be reworked / improved, given steffen's input on the matter