AD: UA: no linearization for UA states that are off

[ebranlard]

This pull request is ready to be merged

Feature or improvement description
This pull request:

• Uses a 2nd order backward difference Newton scheme to perform the time integration of the UA states (should be more robust)
• Does not select UA states that are off for the linearization (results in less states and nicer A matrices (less zeros))

Also:

• Performs an extrap interp of inputs before updating the states, and perform checks on the inputs to make sure they are consistent and dont introduce division by zero
• Uses smaller increments for the linearization of UA continuous states (dx)
• The python script for comparison of lin files aborts only after checking all the lin files, and therfore displays all the errors that occurs (and on which module lin file).

Impacted areas of the software
AeroDyn, UnsteadyAerodynamics

Additional supporting information

Test results cahnge

• OpenFASTLin: Fake5MW*: the size of the states matrices is now changed. The change of dx slightly affects the lin results as well. No changes in frequencies and damping.