Make loading of OpenNLP model location configurable

[chrismattmann]

Don't aways assume there is a model/* root level dir. Load from classpath.

[smadha]

@chrismattmann - #12 should fix this, I tested it yesterday. Will work on Tika to initialise an instance of AgePredicterLocal with additional parameters. Or do you think searching in classpath for model names is better ?

[smadha]

If we do it classpath way either we do something like

SentenceTokenizer.class.getResource("en-pos-maxent.bin")

and keep en-pos-maxent.bin in classpath at same package declaration as SentenceTokenizer which is opennlp.tools.tokenize
Or we can make a fully qualified package exclusively for storing models like opennlp.model and make

SentenceTokenizer.class.getResource("/opennlp/model/en-pos-maxent.bin")

Either way we'll have to keep model at a specific path, so I think making it configurable in AgePredicter is probably better. We can find classpath in Tika like we do now. What say?