Conflicts:
	src/OpenFOAM/matrices/simpleMatrix/simpleMatrix.C
	src/OpenFOAM/matrices/simpleMatrix/simpleMatrix.H
	src/mesh/blockMesh/curvedEdges/BSpline.C
	src/mesh/blockMesh/curvedEdges/spline.C
	src/mesh/blockMesh/curvedEdges/spline.H

- Also adjusted chemistrySolver/EulerImplicit/EulerImplicit.C
  to use simpleMatrix with initialized values

- the element of least surprise

This is controlled by providing a list of the fields to be cached in the "cache"
sub-dictionary of fvSolution.

Debug information about the caching is printed when the solution::debug switch
is on.

There are still a couple of issues to do with the naming of gradients used in
corrected snGrads and limited interpolation schemes that need to be resolved
but these are no different to previously and hence not urgent.

- why doesn't simpleMatrix always do this?

- also, minor doxygen fixup

    deltaT().value() to deltaTValue()
and
    deltaT0().value() to deltaT0Value()
across the whole code - faster to return especially if being used
often, in each call to a submodel for example.

and other unit conversion functions as they are required e.g. slug <-> kg.

by providing the independent objects directly.

Devirtualised functions which should not be overridden and so than some functions can be inlined for efficiency.

http://www.cfd-online.com/Forums/openfoam-bugs/69141-compilation-order-single-precision-issues.html