merged with default

MIN: Changes in GAMIC hdf reader.
Changes include corrections for azimuthal and elevation alignment
of data and missing dynamic range attributes

wradlib/wradlib mit default zusammengeführt

DOC: Corrected typos with respect to freqencies of typical radar bandwidths.

wradlib/wradlib mit default zusammengeführt

ENH: Added funtion read_generic_hdf5.

As discussed in Ticket #11, adding support for reading
all types of hdf5 files.

MIN: Changes on linear_despeckle and fill_phidp.

For linear_despeckle, the first gate is set to nan if the second is nan.

For fill_phidp, the computation has been made more robust for the margins of the arrays.

FIX: Made sure that the setup_distribute module is packaged for PyPI.

MIN: Using the median instead of the mean to compute the value at the margins surronding a NaN region in function fill_phidp.

This way, local jumps in PhiDP do not affect the local fill value too much.

MIN: Removing some redundant ines from fill_phidp.

ENH: Changed the way PhiDP values are filled in function fill_phidp in order to avoid Kdp artefacts.

Citing from the new function docstring: Contiguous NaN regions are filled by the average
of the values of the margins that surround the NaN region. At the left and...

FIX: Removed error in addressing the variable kdp in function kdp2r.

DOC: Enhancing the documentation of the kdp_from_phidp function.

MIN: Making the use of "mean" the default to compute MFB, and "median" optional.

Using AdjustMFB was using the median to compute the mean field bias in order
to make the correction more robust against outliers. In order to account for the
name (MEAN field bias), the behaviour was changed. Now, the field bias is, by default,...

MIN: Considering the possibility that ErrorMetrics does not receive any valid obs/est pair.

REF: xvalidate method in adjust now returns NaN values for invalid data points.

In the previous version, the xvalidate method only returned pairs of observed and
estimated if both observed and estimated were valid values. This had a major
disadvantage: You could not relate the values of the returned obs/est pairs to...

FIX: Added "try" in __init__ so import does not raise an error if speedup is not available.

FIX: Forcing fill_phidp to continue iteration after setting all gates to zero.

MIN: Relaxed conditions on the shape of input arrays for classify_echo_fuzzy.

Now the shapes of the arrays do not need to have the same shape, but only
need to be broadcastable (the two last dimensions need to be the same).

FIX: Removing a bunch of nasty bugs from the latest enhancements related to polarimetry.