Submitted version including Maik's additions.

Added short absract for ERAD2012

Made all correctAttenuation functions behave the same for attenuation overflows

Adjusted overflow handling so that it fits the documentation.

a few readability and logical corrections.
- k is defined as np.zeros, so there is no need to zero out any subset of k
directly afterwards.
- If there are *any* NaNs in the gateset we cannot continue. So let's say that in
the code and throw a proper exception instead of returning something useless....

changed class.rst which should enable *some* kind of class documentation.
Seems to work although it generates a host of strange error messages during
build.

Fixed some documentation bugs

added **kwargs to polar_plot
This way additional keywords can be passed to pcolormesh like vmin and vmax.

removed jsmath-part.
jsmath is a large download and needs its fonts installed on the *user's* system,
which most people won't do just to read our documentation.
So I'll abandon this road for now and stick with LaTeX.

changed dtype of output array to float32 in ipol.py Idw

minor edits in ipol docstrings

Temporarily added the current dependencies to pyproj and basemap until we will
have found an agreement on the dependy policy.

import of utils in __init__.py

added util to __init__.py

minor edit in util.py docstrings

completing function aggregate_in_time
several minro edits in docstrings of georef.py

reworked togrid so that multidimensional data will be handled correctly.
Now more than one image may be gridded simultaneously.

added new moduile util.py

added convenience function projected_bincoords_from_radarspecs in georef.py

the check if the input array is nan, now don't ends in an exception raise, but in a Zero-attenuation array in the output
in the next version i will check if it is better to return a nan array for attenuation