Separate simulator creation from hardware creation

[wtgee]

The various create_<hardware>_from_config scripts are currently also capable of creating a "simulator" version of the hardware. To me that seems non-logical as creating from config should necessarily be creating real hardware.

Describe the solution you'd like
I would propose we remove all simulator creation from these convenience functions and handle simulator creation separately. This could look like a create_from_config (one name for each type of hardware, specified as a class method on the base class) and create_simulator.

Additional context
I think this would help with testing, both for software testing as well as real hardware testing.

Related Issues/PRs
#848 #636 #849 #478 #29 #842 #843 probably more.

Thoughts? @jamessynge @ImmortalRabbit @AnthonyHorton

[jamessynge]

I'm concerned about this proposal because I don't want the production code to have branches such as "if should use simulator". I'm wondering if we can't have a third method: create__driver, which is used by the production code, and is responsible for deciding if we're using a hardware driver simulator or a real driver (possibly with simulated hardware).

I discussed the above with @ImmortalRabbit this morning, and noted that this would mean we'd have a repeated pattern for each type of hardware. This lead me to wonder about some common approach to this, such as an application of the Registry pattern. Thoughts?