update broken links, wrap stdout for python 2 and warning for character-level models

[SwordYork]

Hi, I have provided some fixes.

1. Update the broken links of NMT and optimize_loss.
2. Wrap the stdout for python 2. Maybe a more graceful way for PR Fix unicode print error #96.
3. Add a warning for character-level models. As Handling special characters for character seq2seq model  #153 illustrated, we shouldn't use tf.string_split for unicode strings. However, I could not find any elegant way to deal with it. Maybe we need to implement a new op, for example, tf.decode_unicode(). At least, we should add a warning for this issue.

Thanks.

[googlebot]

Thanks for your pull request. It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

📝 Please visit https://cla.developers.google.com/ to sign.

Once you've signed, please reply here (e.g. I signed it!) and we'll verify. Thanks.

---

• If you've already signed a CLA, it's possible we don't have your GitHub username or you're using a different email address. Check your existing CLA data and verify that your email is set on your git commits.
• If you signed the CLA as a corporation, please let us know the company's name.

[SwordYork]

I signed it!

[googlebot]

CLAs look good, thanks!

[SwordYork]

I have undo the push to github previously.
I don't why ci/circleci does not start.
I have run unit test under python 2 and python 3 locally, also checked with the pylint.

[dennybritz]

Hm, it's weird that CircleCI doesn't run it. I'll look into it and run it manually. Thanks!