`JSONFormat` shouldn't depend on Content-Type to avoid generating invalid JSON

从 src/cloudevents/core/formats/json.py 中引用的行开始，检查 JSONFormatter 对 data 和 Content-Type 的处理。验证字典数据即使在 application/octet-stream 下也能生成有效的 JSON，并运行相关的 formatter 测试或为此情况添加覆盖率。

Hi,
i just upgraded my service to v2.

The problem:

• the service which produces the cloudevents is maintained by another team, i.e. not in my control
• they send the events with Content-Type application/octet-stream
• therefore, in my tests i also emulate sending the events with that Content-Type:

event = CloudEvent(
	attributes={..., "datacontenttype": "application/octet-stream"},
	data=payload,  # <- this is a dictionary
)
message = tobinary_event(event)
client.post("/trigger", content=message.body, headers=message.headers)

now my tests started to send invalid json data (using single quotes instead of double quotes). The reason is this line. E.g. it converts {"asdf": True} to b"{'asdf', True}".

i initially didn't realise this comes from my test case and assumed the SDK generates the invalid data. It took me a while to figure out that only my test setup is wrong.

Proposed solution:

I think str(data).encode("utf-8") is a bad default. Instead, the JSONFormatter should always try to return dumps(data, cls=_JSONEncoderWithDatetime).encode("utf-8"), regardless of the content type. it's the JSONFormatter, so when i pass in a dictionary, i expect it to generate valid json. If it can't, then it's fine to fail with a json.JSONDecodeError or the like. Imo, failing is better than silently generating invalid request data.

Wdyt?