Move calls to GetNewUnit into file opening routines and wrap with !$OMP critical

[andrew-platt]

This is ready to merge

Feature or improvement description
We have had issues with unit numbers when opening files for many years when OpenMP is used. The root issue was that the GetNewUnit call was separate from the actual file opening. This could lead to race conditions when multiple threads hit a GetNewUnit before the retrieved unit numbers were used to open the corresponding files.

To solve this, we moved all the GetNewUnit calls into the Open*File subroutines in NWTC_IO.f90. Inside those routines we wrap the GetNewUnit and open(*) command pairs in a !$OMP critical block. This should prevent the same unit number from being given out to two different OpenMP threads at once.

We debated if this should go into 3.5.5 or 4.0.0. In the end we decided this would be better to address in 3.5.5, mostly so that I don't have to field any more questions or bug reports about it when users stay on the 3.5.x series longer.

Related issue, if one exists
#2510
There were many others over the years that we suspect may be related.

Impacted areas of the software
This change should be completely transparent to the end users, but should fix issues with conflicting unit numbers when OpenMP is used.

Additional supporting information
We originally thought about wrapping the original GetNewUnit calls and the following Open*File calls, but decided that would complicate the code considerably. Instead we decided it was simpler to move the GetNewUnit calls and handle the !$OMP critical block in the library so future programmers would not need to think about it.

Test results, if applicable
No test results should change.

[bjonkman]

For backward compatibility, could you wrap this call to GetNewUnit (here and the other places it's now called) in an IF statement like IF (Un < 0) call GetNewUnit()?

[bjonkman]

Alternatively, GetNewUnit could be changed to start looking at units at the given Un value if it's >= 0. Otherwise, Un is really intent(out) only.

[andrew-platt]

I'll wrap those calls in an if statement. I think we want to maintain that backwards compatibility for now.

[andrew-platt]

After discussing this with @deslaughter, we concluded that there are to many code changes going into this PR to make sense for a 3.5.5 release.

I am closing this PR and will open a new one to dev with a reworked version of these changes.