Optional fields in Python dataclasses

I think optional fields should be optional in generated Python dataclasses.

[Demo link][1]. Source language: TypeScript

```ts
export interface Config {
a?: any;
b?: any;
}
```

Target language: Python 3.7 (abridged for clarity)

```py
from dataclasses import dataclass
from typing import Any

@dataclass
class Config:
a: Any
b: Any
```

The problem here is that, if you try to instantiate a `Config` object, the fields are both required:

```
>>> Config(a=12)
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
in
----> 1 Config(a=12)

TypeError: __init__() missing 1 required positional argument: 'b'
```

Since the properties are both optional, this should be OK. The way to make that happen with a dataclass would be to generate code like this:

```py
@dataclass
class Config:
a: Any = None
b: Any = None
```

Then I can construct a `Config` omitting either property:

```
>>> Config(a=12)
Config(a=12, b=None)
```

I realize you get the expected behavior if you use the generated `config_from_dict` method. But you also get worse type checking (via your linter) for `dict`s than for `dataclass`es with named parameters.

Thanks for the great tool!

[1]: https://app.quicktype.io?share=XBPvukZ1HNzpDIioKrRb