eigen3 correction

[karlnapf]

@vigsterkr UPDATE: I applied the fix to develop as it is definitely necessary. The rest is optional for cleaner cmake options.

I renamed them to start with LINALG_ so that they are grouped if ppl check the options and not scattered all over without it being clear that they change linag behaviour. Also, ccmake which shows the cached options now does't show a binary ON/OFF, but the string. Just like CMAKE_BUILD_TYPE

This might interfere with some builds. Let me know

[karlnapf]

HAVE_LINALG was not enabled previously, which was an error

[karlnapf]

The viennaCL buildbot requires to change DLinAlgBackend=VIENNACL to DLINALG_DEFAULT_BACKEND=VIENNACL

[vigsterkr]

@karlnapf any reason why not to merge this?

[karlnapf]

needs a buildbot change, see comment above. I would prefer this.
Not important as it is refactored anyways. Merge if it is easy to change buildbot scripts, otherwise just close

[vigsterkr]

i've merged and updated the buildbot params for the viennacl job, i'm getting the following output from cmake

-- ViennaCL set as default global linear algebra backend library
-- Eigen3 set as default core module (linalg) backend library
-- Eigen3 set as default reduction module (linalg) backend library
-- Eigen3 set as default eigen solver module (linalg) backend library

@karlnapf why Eigen3 is still the eigen solver module?

[karlnapf]

This has to be fixed

[karlnapf]

I guess this can wait until we re-structure no?