Handle Unicode escapes

Strings in JSON can contain escaped Unicode characters (i.e. \uXXXX) and the spec says that it's legal for any character to be escaped (even though technically you only need to escape a few of them).

The Java LSP client library (lsp4j) is particularly aggressive about escaping characters, so it immediately runs into this issue. For example, it sends:

```
{
"jsonrpc": "2.0",
"method": "textDocument/didOpen",
"params": {
"textDocument": {
"uri": "file:///Users/mdb/ops/langs/reason/a-reason-react-tutorial/src/Main.re",
"languageId": "re",
"version": 1,
"text": "ReactDOMRe.renderToElementWithId(\u003cTodoApp /\u003e, \"root\");\n"
}
}
}
```

(which escapes < and >) and this fails thusly:

```
Scanf.Scan_failure(\"scanf: bad input at char number 35: \\\"illegal escape character 'u'\\\"\")
```

I hacked the code to turn \u00XX into \xXX just to work around this for now (it would still fail on any non-ASCII escaped Unicode characters). I don't name my variables with UTF-8 smileys. 😁 But a proper decoding of Unicode escapes will be needed at some point.

No source file or test is named in the issue. Start by locating the JSON string escape decoder and reproduce the lsp4j example, including \u003c, \u003e, and a non-ASCII escape; done means valid \uXXXX escapes are decoded instead of rejected.

Written by the indexing model from the issue text.