Fixes to LU format for phraselist handling as well introduction of new parser flags for collision handling for labels

[vishwacsena]

• Includes fix to an issue where imported phraselists were not converted to LUIS JSON correctly. Old output still had "modelName": "phraseListName" which would not work and this instead needs to be "featureName": "phraseListName".
• Add new capability to LU format to signal parser how to handle collisions for entity labels.

> !# @parser.OnAmbiguousLabels = takeFirst | takeLast | takeLongestLabel (default) | throwAnError

takeFirst will accept the first label. As an example, the first label vishwac for userName will be accepted.

> !# @parser.OnAmbiguousLabels = takeFirst

# intent1
- my name is {@userName = vishwac} sena

# intent1
- my name is {@userName = vishwac sena}

takeLast will accept the last label in case of collision. As an example, the last label vishwac sena for userName will be accepted.

> !# @parser.OnAmbiguousLabels = takeLast

# intent1
- my name is {@userName = vishwac} sena

# intent1
- my name is {@userName = vishwac sena}

takeLongestLabel will accept vishwac sena since that is the longest label for userName

> !# @parser.OnAmbiguousLabels = takeLongestLabel

# intent1
- my name is {@userName = vishwac} sena

# intent1
- my name is {@userName = vishwac sena}

throwAnError will throw an exception when duplicate overlapping labels are found for the utterance.