spate-avro 0.1.0

Avro deserialization for the Spate framework: Confluent wire format, schema registry integration, and schema-evolution support. Applications should depend on the `spate` facade crate with the `avro` feature.
Documentation
# spate-avro

Avro deserialization for the
[Spate](https://github.com/spate-etl/spate) framework: bare-datum
decoding for Confluent wire format (magic byte + schema id), Avro
single-object encoding, and raw datums with a fixed schema — into
dynamically-typed `AvroValue` records or your own serde types.

Registry schemas are fetched by an async task on the I/O runtime and
cached per process. A cache miss never blocks a pipeline thread and never
drops the record: the batch reports "not ready" and replays once the
schema lands. Unresolvable ids are negatively cached with a TTL and then
follow the deserializer's error policy like any poison payload.

Schema evolution follows Avro resolution rules via an optional
`reader_schema` (field reordering, defaults, promotions, aliases).

Key types: `AvroDeserializerBuilder` (from the
`deserializer: { avro: ... }` section), `AvroSerdeDeserializer<T>`,
`AvroValueDeserializer`, and — behind the `fast` feature —
`AvroSerdeDeserializer<T>`.