Fix Failing/Passing SplineInterpolation test

[sslattery]

The SplineInterpolation test fails on my machine but passes on most of the CDash builds. There are CDash builds in which this test passes some nights and not others. The test started failing after Tpetra switched to view semantics as part of the Kokkos refactor. This issue will track the process of finding the bugs. It may be a memory leak.

[sslattery]

valgrind seems pretty clean. I don't see any invalid reads or writes. messing around with the thyra operators gives some weird results. Maybe thyra wrappers for tpetra were not appropriately updated?

[sslattery]

verified that on apply the target vector is being overwritten with zeros

[sslattery]

After some work on this it is still not clear what the issue is. Although they involve a global linear solve at every apply of the operator, the splines are desirable because they are less sensitive to the choice of support radius. Still looking into this.

[sslattery]

It was identified that a bug in Tpetra caused this. Depending on the Trilinos version, using beta=0 for a matrix vector multiply (the default) caused the vector to be filled with zeros. The current 11.14 patch tarball passes in release and fails in debug while the current state of the public repository passes for both. Since this is passing using the current state of Trilinos, this is fixed.