Fix tokenizer insert empty token to sentence object

[eurekaqq]

Fix issue#1188
This issue is about tokenizer of segtok library.
In segtok, someword+ n't, e.g. don't, didn't...
It will be tokenized as someword and n't.
However, if user has split it, it will insert empty between someword and n't.
Example:
"do n't" -> "do", "", "n't"
"did n't" -> "did", "", "n't"

To Reproduce

from flair.data import segtok_tokenizer

text = r'do n’t'
tokens = segtok_tokenizer(text)
print(tokens) #[Token: do, Token: , Token: n’t]

This commit fix this problem.

[alanakbik]

Thanks! The travis error is from a different commit, now fixed, so will merge!