SystemExit and sys.exit behavior overlap

[thecnoNSMB]

Rule request

raise SystemExit(arg) and sys.exit(arg) are identical in behavior, one should be favored over the other for consistency.

Reasoning

I don't know which of the two would be preferable for this linter (I suspect sys.exit(arg), but raise SystemExit(arg) doesn't require any imports), but several other contexts where multiple forms are valid have a rule enforcing one form as the standard (such as single vs double quotes, or string formatting methods), and I figure the same logic should apply here.

[thecnoNSMB]

quit(arg) and exit(arg) should also be prohibited under this rule, as those are from the site module and are only intended to be used for the REPL; as such, they may not be loaded at all when a script runs.

[orsinium]

I suspect sys.exit(arg), but raise SystemExit(arg) doesn't require any imports

Yes, I am for sys.exit.

1. Importing of sys is incredibly fast, I'd even say free:

   import sys
   sys.__loader__  # _frozen_importlib.BuiltinImporter
   sys.meta_path.index(sys.__loader__)  # 0

2. It should be clear that this statement isn't an exception that can be caught somewhere up in the stack but an uncompromised exit without turning back.

quit(arg) and exit(arg) should also be prohibited under this rule

WPS already has this one:

WPS421 | Found wrong function call: exit

However, I guess it should be more verbose like in pylint:

R1722 | Consider using sys.exit()