Implement python source model reader

[Anton Gritsay]

The code from modelist toolkit can be used: https://github.com/gem/mtoolkit/blob/master/mtoolkit/smodel.py

The following files should be got rid of:

java/org/gem/engine/LogicTreeProcessor.java
java/org/gem/engine/GmpeLogicTreeData.java
java/org/gem/engine/hazard/parsers/SourceModelReader.java

The important changes should be introduced to logic tree processing architecture then: transition from GR MFDs specified as {a, b, mMin, mMax} to evenly discretized function should be delayed as long as possible, i.e. up to the computation execution.

Currently source model reader converts {a, b, mMin, mMax} to a set of precalculated values (loosing precision), then in order to apply uncertainty we have to convert those values back to {a, b, mMin, mMax} (loosing precision) and then back to evenly discretized function again (guess what? loosing precision!) Then the result histogram is sent to a worker.

Instead we should make source model reader keep origin values {a, b, mMin, mMax}, apply uncertainties to them being unchanged and send those four values to workers and make them create the histogram if they need so.

[Lars Butler (lars-butler)]

We have this now.