[imikejackson]

Summary

Three Orientation Analysis filters created geometries with a physical length unit although their coordinates are not lengths:

Filter	Geometry	Coordinates	Old unit
Convert Orientations To Rodrigues Geometry	Vertex geometry	Dimensionless orientation-space points	Meter
Generate and Write Pole Figure Images	Composed image and count image	Pixel indices	Micrometer
Compute GBCD Pole Figure	Stereographic image on [-1,1] x [-1,1]	Projection values	Micrometer

All three now create their geometries with IGeometry::LengthUnit::Unspecified. The numeric coordinates are not changed.

• Add an optional IGeometry::LengthUnit units argument to both CreateVertexGeometryAction constructors, in the same style as CreateImageGeometryAction. The default is the previous Meter value, so the other callers of the action are not changed. The unit is applied in apply(), is kept by clone(), and has a units() getter.
• Make CreateImageGeometryAction::clone() pass its unit to the copy. The copy used the default Micrometer value before.
• Add unit assertions to ConvertOrientationsToVertexGeometryTest (with and without fundamental-zone conversion), WritePoleFigureTest (composed image and count image), and ComputeGBCDPoleFigureTest. Before the fix these reported 10 == 100 (Meter) and 6 == 100 (Micrometer).
• Add a test for CreateVertexGeometryAction that covers the four array handling modes, preflight and execute modes, direct and cloned application, and the default unit.
• Add one or two short sentences to the three filter documentation pages that explain the unit choice.

Fixes #1803

Test Plan

• New unit assertions fail before the fix and pass after it
• ctest -R "ConvertOrientationsToVertexGeometry|WritePoleFigure|GBCDPoleFigure|CreateVertexGeometry|CreateImageGeometry|ImageGeom" passes (87 tests, serial run, macOS arm64 Release)
• Full tree builds after the core action header change
• clang-format 19.1.1 applied to the touched files

[joeykleingers]

Fix the last finding and otherwise this looks good.

[joeykleingers]

Approving at ff75b2eb4. I pushed that commit to fix the last review findings:

• CreateVertexGeometryAction checks the geometry it creates before it sets the units. When another DataObject already uses the path, it returns error -6108 with a fix hint instead of dereferencing a null geometry. A new GeometryTest case covers this in preflight and execute.
• The new pole-figure unit checks use REQUIRE_NOTHROW on each geometry lookup first, and the new test pointers end in Ptr.

Verified locally on macOS arm64 (simplnx-Rel): the build has no warnings, and the 36 affected tests pass. Without the null check, the new test fails with SIGSEGV.

Problems that existed before this PR are not fixed here. I am filing or linking issues for them.