Dask-geomodeling

------------------

- Reverted changes to elemwise.

------------------

- Added raster.Max block to select the maximum value from overlapping sources.

- Made Elemwise blocks more strict regarding source periods.

- Drop python 3.5 support and move on other version requirements.

- Fix deprecation warning with distutils.

------------------

- Added a default setting "raster-limit-timesteps".

------------------

- Fix field_operations.Classify if used with int-typed labels. A NaN value in
the result resulted in a Categorical output dtype. To fix this, Classify
now returns floats also when input labels are integers. (82)

------------------

- Make Dilate arguments JSON serializable. (81)

------------------

- Never return Categorical dtypes in the field_operations.Classify and
ClassifyFromColumns. This leads to pandas incompatibilities with later
operations (round, subtract, where, mask).

- Never return Categorical dtypes from ParseTextColumn. (79)

- field_operations.Where and field_operations.Mask now also allow non-boolean
conditionals. This was already the case for Where on pandas == 0.19.*. (78)