Better debugging for exec / json parsing

[rtreffer]

While debugging the exec issue in #248 I've mostly been greeted by generic JSON parse errors.
I would have expected some debug output about the actual issue.

Examples of failing code paths:

• lnnode.py

    def getnewaddress(self):
        res = json.loads(self.lncli("newaddress p2wkh"))
        return res["address"]

• ln_test.py

inv = json.loads(base.warcli("lncli 2 addinvoice --amt=1234"))["payment_request"]

To improve debugging it would be great to have a "execute + parse json" method that

1. runs the command
2. tries to parse the output
3. on error: log the command + output and re-throw an error

Alternately it could be added to all code paths, but given that json output is the most common output I would expect it to be very redundant.

[josibake]

Seems like we could use a decorator for this, similar to @exponential_backoff. What I'm envisioning is a pattern where functions return un-parsed outputs and then we add a decorator for parsing the output and catching any errors. We could add specific decorators for different output types, but starting with a @parse_json_response decorator seems like a good start.