client, handle when source contains `\u##` in doc

[weidongxu-microsoft]

This likely result in “illegal Unicode escape”. A typical case be domain\\username in Swagger description.

Some step in the chain (I think it be m4 / tsp compiler) would make it domain\username in code-model.yaml

Then \u## is now parsed as escape for Unicode, which typically fails Java compiler.

---

Should we encode it back to e.g. domain\\username?

(checking every "description" text for this could affect performance)