Because the program takes in multiple files, it seemed like to me that parallelizing would yield some performance benefits. I didn't have time to sort the resulting hashmap, so thats the only requirement I missed.
To run the unittests:
$ mvn clean test
To create a commandline executable
$ mvn assembly:assembly
From there you can see the help menu by running:
$ java -jar target/wordcount-1.0-SNAPSHOT-jar-with-dependencies.jar
-f Path to the file with list of filenames
-h show help.
-t Set the number of thread to use.