Use a parsed schema in the cached decoder function

[0x26res]

Description

MessageSerializer#_get_decoder_func should call fastavro#parse_schema to improve performance.

There is a minor overhead to this function call (parse_schema), so it would be beneficial if it was only done once, and the parsed schema would be binded to schemaless_reader when creating the lambda that constitutes the cached decoder function

Later, the call to parse_schema within schemaless_reader is smart enough to figure out that the schema is already parsed (https://github.com/fastavro/fastavro/blob/master/fastavro/_schema_py.py#L83)

How to reproduce

I'm not sure if it's applicable as it is not a bug. This is covered by TestMessageSerializer either.

Checklist

Please provide the following information:

• confluent-kafka-python and librdkafka version (confluent_kafka.version() and confluent_kafka.libversion()): 0.11.4
• Apache Kafka broker version: 0.11.0.3
• [ ] Client configuration: {...}
• Operating system: Linux
• Provide client logs (with 'debug': '..' as necessary)
• Provide broker log excerpts
• Critical issue

[0x26res]

This looks like it's done:
#627