docstrings: updated!
costfunction: CFAdapterScipyCG obsolete, deleted estimate_lambda()
dataset: use_fftw and threads obsolete
diagnostics: implemented stuff... much stuff...
fft: Docstrings and implementation of threads
magcreator: fixed bug in slab (dimension mix-up)
phasemap: plot now also returns colorbar
reconstruction: PrintIterator and optimize_simple_leastsq obsolete
regularisator: Docstrings
scripts: minor modifications
scripts - reconstruct_random_pixels: now use optimize_linear
testcases: analytic, compliance, kernel, magcreator, magdata, projector

kernel, phasemapper: now use the new pyramid.fft module.
MagData, PhaseMap: now use np.float32 instead of np.float64
MagData: new quiver_plot3d with just matplotlib (buggy and still WOP).
logging: deleted several logging.debug prompts for speed up.
util: deleted module!

Se_inv: is now handled differently, defaults to eye-matrix, but can be set
        later (added a few convenience functions to DataSet)
utility: (re-)added utility module for index conversion
kernel/phasemapper: can now use FFTW library and operate on pre-allocated
        arrays
dtype: set dtype of arrays to np.float32/np.complex64 (halves memory size),
       also applied to numcore routines!
magdata: quiver_plot can now plot every n-th arrow (yet untested!)
reconstruction scripts: modified several reconstruction scripts.

setup: Changes to setup.py, pyramid.version to fit to numpy and jutil
       now uses __all__ magic and supports from pyramid import *
tests: changes to test_compliance/magcreator/analytic (should now work)
collaborations: added several scripts for Patrick
reconstruction: added batch file for the reconstruction of several configs
test methods: (unfinished) scripts for fftw tests and kernel comparisons
paper 1: ch5-4 changed padding values to 0, 1, 3 and 7
plotting: deleted plt.show() lines (have to be done manually now or via Spyder)
logging: now named _log instead of LOG (private variable instead of constant)
         also some logs were disabled, because Pylint says they are slow...
reconstruction: now has info argument (pass a list in which cost info is copied)
phasemapper: changed sign of constant PHI_0, is now correctly positive (instead
             the negative z-integration is taken into account which was not done
             before!)
kernel: now uses FFTW, numpy code was changed accordingly and is still available
phasemapper: also uses FFTW now, work in progress: FFTW for jac_T_dot