Make LinearSystemId functions return LinearSystem with fully observable state

[calcmogul]

LinearSystemId.IdentifyPositionSystem() returns a LinearSystem<2, 1, 1>, which is incompatible with the physics simulation classes.

Instead, it should return a LinearSystem<2, 1, 2>, and LinearSystem should have a function for slicing the states, inputs, and outputs so that it can produce a LinearSystem<2, 1, 1> for use with a Kalman filter.

[narmstro2020]

I've wanted this fix for a bit myself. Seemed like a weird gap after reading all of the documentation on the subject.
I can take a whack at it. This should also apply to createElevatorSystem and createSingleJointedArmSystem. Flywheel, DCMotor, and the DriveTrain should be left alone if I'm not mistaken?

[calcmogul]

Yea, those three already do the correct thing.

[narmstro2020]

On a marginally related subject I'd also like to take the type to add a WheelSim class that simulates something like a drive wheel. It would effectively be the same as DCMotorSim, but reading out linear units instead of angular ones.

[calcmogul]

When would that be used? Seems like a lot of API/library overhead for a single scalar multiplication on the user side.

[narmstro2020]

My use case would be in terms of simulating a drive wheel given starting with kV and kA values in VoltsPerMPS instead of RadPerSec.

Perhaps a constructor that takes in kV and kA in the linear units along with a conversion factor in the existing DCMotorSim class?

[calcmogul]

A constructor overload like that is a non-starter in C++ because it isn't unit-safe.

I'm also hesitant about a WheelSim class because it bloats the API with overlapping functionality.

[narmstro2020]

Okay. I'm content without. Just an idea.