feat!: support string-encoded json

[daniel-sanche]

We recently merged a PR into the v3.0.0 staging branch adding support for logging dicts as jsonPayload logs in our custom handlers:

logging.error({'a':'b'})
↓
{"a" : "b"}

While this can be useful in basic logging set-ups, the Python logging API expects string-like msg inputs, so this can't always be relied on to work properly in more complex logging environments.

With this PR, the handlers will also attempt to parse a dictionary out of the final log string before sending it on Cloud Logging, so string-encoded dictionaries will work as expected:

logging.error('{"a": "b"}') 
↓
{"a" : "b"}

This also allows you to set a formatter object with logging attributes once, and then continue to log strings as normal. Each log will now show up in jsonPayload format in CloudLogging

formatter = logging.Formatter(fmt='{"thread": "%(threadName)s", "message": "%(msg)s", "hardcoded": "test" }')
handler.setFormatter(logFormatter)
logging.error("hello")
↓
{"message": "hello", "thread": "MainThread", "hardcoded": "test" }

You can also specify in custom fields using the extra argument:

formatter = logging.Formatter(fmt='{"message": "%(msg)s", "custom": "%(custom_arg)s" }')
handler.setFormatter(logFormatter)
logging.error("hello", extra={"custom_arg":"test"})
↓
{"message": "hello", "custom": "test" }

Fixes #331