fixes issue #1, also saves alignments to a file

Closes issue #1 
Saves alignments to file

See merge request !1

 - DAG nodes can be single-word constituents (heads of pred/arg constituents)
 - Had to initialise weight vector with 1s to make the inference work
Updated to lbjava-maven-plugin

Added LLM scorer
Added SRL predicate sense as an attribute
Added evaluation code in Main
Added JUnit test
Converted prepSRL objects to phrase-based constituents
DAG now contains only phrase-level nodes

 - Removed higher-level semantic DAGs
 - Removed StructureFeatureGenerator
 - Removed "uncovered tokens" (dep-based)
Updated dependencies

Commented out the dependency-generated DAG parts
 - Prep/nom SRL produces error reduction on selected phenomena
Added the last 2 files from the "ignored" data folder
Minor renaming/bug fixing

Commented out most of the RATER code
Added prep/nom SRL
Removed TEModelBaseScorer (replicated its methods in place)
Slightly changed the way constituent SRL attributes are generated
Added some working notes as TODOS
Prettified code:
  -renaming variables
  -removed unnecessary exceptions
  -added comments

Added the previously ignored data and cache source code folders.
Changed the Dependency annotator to the HeadFinderViewGenerator, based off Charniak since it seems to be much better at finding usefull dependencies.
Removed Quantities annotation until its problems are fixed
Added markdown README
Annotator unit test now depends directly on Annotator
Minor cleanup, renaming of variables