Figure out configuration options for stateful machines

[Liam-DeVoe]

Hypothesis puts configuration for stateful machines, like stateful_step_count, on @settings. This makes sense in Hypothesis because state machines are top-level tests. In Hegel, a single test case of a state machine is invoked by hegel::stateful::run inside of a hegel::test, so it's less clear where to place our stateful_step_count equivalent.

Options:

• (1) Add to #[hegel::test] settings. Applies to all state machines invoke in that test.
• (2) Add settings parameters to hegel::stateful::run. Applies to just that test case.
  • This implies stateful_step_count can vary between test cases. Are we okay with this? Will this cause flakiness?
• (3) Add settings to #[hegel::state_machine]. Applies to that state machine for all test cases.

I think my preference is (3) > (2) > (1).

See also #148 (comment).

[DRMacIver]

I think my preference is some hybrid of (2)/(3), but as proposed I share your (3) > (2) > (1) preferences pretty strongly.

I think possibly we should give state-machines a configuration operation and write it as something like my_machine.configure(|c| c.max_steps(100)).run(tc) ?

Hmm. I guess we currently have hegel::stateful::run(machine, tc). Maybe we shouldn't and that should be a method defined on the machine by our proc-macro?

I think it's OK for stateful_step_count (max_steps?) to vary between test cases and can't cause flakiness as long as it varies according to the usual rules for Hegel determinism which is that the maximum number of steps depends only on generated data.

[Liam-DeVoe]

I think we have several basically-fine options here. I think adding settings to #[hegel::state_machine] is basically-fine. I think changing hegel::stateful::run(machine, tc) to a builder and configuring settings there is also basically-fine.

I think I have a slight preference that my_machine.run(tc) not exist, because tc or the hegel module should be positioned as the origin of control, not the state machine.