[test] for coreNLP: allow to use the already started corenlp

Thing is: I sometimes have corenlp/postgres running for my dev
env. When I start tests, they fail because they try to start corenlp
again on the same port. So what I added was checking for that "address
already in use" condition and going on with the tests. CoreNLP is
stateless so it shouldn't matter which one we use for tests.
Status Job ID Name Coverage
  Cabal
canceled #18333
cabal

 
  Stack
canceled #18334
stack

 
  Bench
canceled #18335
allowed to fail manual
bench

 
  Test
canceled #18336
test