Preallocate JTC variables to avoid resizing in realtime loops

[AndyZe]

• Preallocate several variables
• last_commanded_state_ was previously used for open-loop control. I think it's redundant with state_desired_ now, so delete it

[erickisos]

LGTM!

[AndyZe]

I don't understand why quite a few unit tests are failing. Marking as draft until that's fixed.

[mechwiz]

I think #296 broke the JTC tests which is why many of them fail.
I think #334 is an attempt to fix the issue but it hasn't yet been resolved.

[AndyZe]

OK, it does look like this PR does not make the test failures worse. Re-marking it as ready for review.

[mergify]

This pull request is in conflict. Could you fix it @AndyZe?

[AndyZe]

The only possibly-controversial thing I see is error handling in case dof_ isn't initialized, and that's just because I didn't see a better way to handle the error there. I could even delete it, if you'd rather.

  if (dof_ == 0)
  {
    fprintf(stderr, "During ros2_control interface configuration, degrees of freedom is not valid;"
      " it should be positive. Actual DOF is %zu\n", dof_);
    std::exit(EXIT_FAILURE);
  }

This could be turned into an assert() if you'd rather.

[destogl]

I find this very good! Thanks! Just one small comment.

[destogl]

can we resize those in on_configure? Since this should be, if possible, real-time capable method.

[AndyZe]

That makes sense. Done and tested.

[destogl]

The failing CI things are OK. Merging this.