Only cmake is supported from now on.

It seemed strange to me to demonstrate a second-order tangent field on a
first-order base function.  People may think that this is a necessary restriction.

Also, this patch includes a bugfix:  The basis of the tangent space was hard-wired
to (0,1,0) and (0,0,1).  But this only works if we are at the tangent space at (1,0,0).

In particular, assemblers now hand LocalView objects around, rather than pairs of
elements and LocalFiniteElements.

[cleanup] Move the FD approximation of the GFE function gradient from the GFE function into the test

It is a part of the test harness, and should not clutter the implementation of the
GFE functions themselves.

Use 1e-8: The square root of the machine precision is said to be the best choice.

Needed for projection-based finite elements.

This patch also adds a unit test.  So far this test passes.

The program is not really a test, but it doesn't really belong into 'src', either.

[[Imported from SVN: r9997]]

[[Imported from SVN: r9913]]

[[Imported from SVN: r9912]]

[[Imported from SVN: r9877]]

That should be no problem now that we use a 50-decimal digits type
for the FD approximation.

[[Imported from SVN: r9876]]

[[Imported from SVN: r9875]]

It offers a 'double' interface, but we don't really need that.

[[Imported from SVN: r9874]]

[[Imported from SVN: r9873]]

[[Imported from SVN: r9872]]

[[Imported from SVN: r9870]]

This code compares scalar and vector AD, and a finite difference
approximation.

[[Imported from SVN: r9869]]

[[Imported from SVN: r9681]]

[[Imported from SVN: r9680]]

[[Imported from SVN: r9614]]

[[Imported from SVN: r9573]]

[[Imported from SVN: r9436]]

This still fails -- there must still be a bug in the Hessian computation

[[Imported from SVN: r9427]]

[[Imported from SVN: r9420]]

[[Imported from SVN: r9419]]

Hence many input configurations are tested now

[[Imported from SVN: r9418]]