Add merge nodes action and two theory manuals, fix rendering and picking defects

Author: tomas-soltys

src/fea-solver/doc/RELEASE_NOTES.md

@@ -13,3 +13,30 @@
   apart. A block whose ratio is one throughout is assembled correctly; a block
   whose ratio is some other number throughout carries that factor too many.
   `RSolverFluid` documents what the check has found so far
+
+### Bug fixes
+
+- The electrostatic element results are densities. The recovered potential
+  gradient was weighted by the Jacobian determinant of the element, and on a
+  surface by its thickness as well, so the electric field, current density,
+  electric energy and Joule heat all scaled with the size of the element they
+  were computed in and drifted instead of converging as the mesh was refined.
+  The electric potential and the electrical resistivity were never affected
+- The **Charge density** boundary condition enters the system with the sign the
+  Poisson equation calls for, on every entity type. It was assembled negatively
+  on line, surface and volume elements and positively on point elements, so a
+  positive space charge lowered the potential around it
+- Joule heat is the dissipation density `sigma*|E|^2` in `W/m^3`, the source
+  density a heat task integrates over the element, rather than that density
+  multiplied by a characteristic element size
+- These corrections change the results of every model using the
+  *Electro-statics* problem type, and the statistics printed for them. The units
+  in the solver log follow: relative permittivity is `N/A` rather than `C^2`,
+  charge density `C/m^3` rather than `C`, electric energy `J/m^3` and Joule heat
+  `W/m^3`
+- The magnetostatic source term uses the vacuum permeability instead of the
+  vacuum permittivity and is assembled in its Galerkin form, and the current
+  density it reads from the electrostatics result is averaged onto the nodes
+  rather than taken from one neighbouring element. Every magnetostatic result
+  changes. The problem type is still incomplete - no boundary condition exists
+  for it, so the level of the computed field is arbitrary

src/fea/CMakeLists.txt

@@ -622,6 +622,7 @@ qt_add_resources(fea icons-geometry
         pixmaps/range-element_create.svg
         pixmaps/range-element_remove.svg
         pixmaps/range-merge_near_nodes.svg
+        pixmaps/range-merge_nodes.svg
         pixmaps/range-move_node.svg
         pixmaps/range-node_remove.svg
         pixmaps/range-node_remove_and_close.svg