Kafka `to_binary()` crashes on events without `datacontenttype` or with non-string attributes

从 cloudevents/v1/kafka/conversion.py 开始，将 to_structured() 与 issue 中提到的较新的 core Kafka binding 进行比较。验证两种复现都能在不抛出异常的情况下完成，在缺少 datacontenttype 时省略 content-type，并将生成的 headers 中的非字符串扩展属性转换为字符串。

Description

to_binary() in cloudevents/v1/kafka/conversion.py raises on two kinds of perfectly valid CloudEvents:

1. Missing datacontenttype (optional attribute) → KeyError

    headers = {}
    if event["datacontenttype"]:                      # KeyError if unset
        headers["content-type"] = event["datacontenttype"].encode("utf-8")

datacontenttype is optional, so event["datacontenttype"] raises KeyError for any event that does not set it. The sibling to_structured() in the same file already guards this with if "datacontenttype" in attrs:.

2. Non-string attribute value → AttributeError

            if value is not None:
                headers["ce_{0}".format(attr)] = value.encode("utf-8")   # AttributeError if not str

CloudEvents extension attributes may be non-string (e.g. Integer/Boolean), so value.encode(...) raises AttributeError: 'int' object has no attribute 'encode'. The newer core Kafka binding already does str(attr_value).encode(...).

Reproduction

from cloudevents.v1.http import CloudEvent
from cloudevents.v1.kafka import to_binary

# 1) no datacontenttype -> KeyError: 'datacontenttype'
to_binary(CloudEvent({"type": "t", "source": "s"}, {"a": 1}))

# 2) non-string extension -> AttributeError: 'int' object has no attribute 'encode'
to_binary(CloudEvent({"type": "t", "source": "s",
                      "datacontenttype": "application/json", "ext1": 5}, {"a": 1}))

Expected behaviour

to_binary() should handle both valid events: omit the content-type header when datacontenttype is absent, and stringify non-string attribute values.