More streamlined support for config objects

[DavidBuchanan314]

In my program, I defined my configuration options as their own dataclass, so that the config could more easily be passed around between functions that need it. I was able to integrate clize with this approach remarkably cleanly:

from dataclasses import dataclass

@dataclass(kw_only=True)
class MyConfig:
	"""
	An example program.

	:param foo: foo docs
	:param bar: bar docs
	""" # this docstring is used by clize

	foo: int = 1
	bar: int = 2

def main(cfg: MyConfig):
	print("running main with cfg:", cfg)

if __name__ == "__main__":
	from clize import Clize
	import sys

	try:
		cli = clize.Clize.get_cli(MyConfig)
		cfg = cli(*sys.argv)
		if type(cfg) is not MyConfig:
			print(cfg)
		else:
			main(cfg)
	except clize.errors.ArgumentError as e:
		print(e)

However, it's a bit unergonomic. I have to import sys myself, and I lose out on the _fix_argv functionality that run() would normally do internally, and I have to implement help/error printing myself too.

I think it would be nice if I could just do this:

if __name__ == "__main__":
	from clize import run
      
	cfg = run(MyConfig, exit=False)
	main(cfg)

For this to happen, run() would have to return ret, in the case that exit is not called explicitly, as opposed to the current behaviour that just returns None.