add option to retain unicode escapes in codegen

Currently, as far as I can tell, codegen always turns unicode escapes into literal unicode characters. 

For instance,

```ts
const foo = "\u{123}"
```
becomes

```ts
const foo = "ģ"
```

This is problematic in cases where we need to the output to be ASCII only, as even if the input is ASCII, after transpiling the result is not ASCII.