Solve Option 2 Performance enhancements

[bjonkman]

Parts of the SolveOption2 algorithm have been moved to the the AdvanceStates routine so that BeamDyn, InflowWind, AeroDyn, and ServoDyn use calculated inputs instead of extrapolations for inputs at t+dt.

This change allows BeamDyn to use calculated rotations and AeroDyn/DBEMT to use calculated velocities, which otherwise would require a correction step. In DBEMT this was a problem with high turbulence and low wind speeds.

[andrew-platt]

After chattting @ashesh2512, we should rerun a test with BeamDyn coupled to OpenFAST (with all forces and degrees of freedom off) and check if the residual is zero once this is merged into dev.

Some external collaborators have noted that the timestep must be really small. @bjonkman added the code in this PR to fix this issue at Envision.

[bjonkman]

As expected, due to the slight changes in inputs, the regression tests fail. However, the results do not look much different than the baseline. These differences should be similar to the what is seen if you add a correction step to the models (particularly the onshore models).

I would appreciate if this could be merged in the near future... it's been waiting more than a year.

[bjonkman]

If it helps, here is the new algorithm for AdvanceStates (it uses calculated inputs when updating states):

[andrew-platt]

@bjonkman updated this doc with the AdvanceStates info: https://github.com/old-NWTC/FAST/blob/master/Docs/TheoryGuide/FAST_Algorithms/FAST_Algorithms_v8.tex

Do we have a good place to migrate some of this info?

[rafmudaf]

We could add a Theory section here: https://raf-openfast.readthedocs.io/en/docs-fast8readme/source/user/openfast/index.html

[bjonkman]

Thanks, @rafmudaf and @andrew-platt!