feat(avro): add --format avsc option to export Avro JSON Schema

[DavidMatzek]

Summary

Adds a --format [idl|avsc] option to the avro exporter (default idl, existing behavior unchanged). With --format avsc, the exporter writes native Avro JSON Schema (.avsc) files directly, one self-contained file per top-level struct. Users no longer need to export .avdl and convert it with avro-tools idl2schemata.

Changes

• --format idl|avsc option (case-insensitive, default idl).
• AVSC renderer reusing the existing struct, enum and dependency-resolution helpers, so both formats describe the same types.
• Records and enums are embedded where first referenced and referenced by bare name afterwards, since Avro allows a named type to be defined only once.
• Every field is a ["null", T] union, matching the IDL union { null, T }.
• Enums list UNKNOWN first and set "default": "UNKNOWN" (IDL: = UNKNOWN).
• --include-array-record produces a <Struct>Array root record wrapping the struct.
• --file-prefix applies to .avsc filenames.

Documentation

• docs/avro.md: new "Output formats" section (IDL vs AVSC comparison, AVSC embedding rules, examples) and --format in the options table.
• CHANGELOG.md: entry under Unreleased.

Tests

• Unit tests for the AVSC generator (fields, enums, nested and referenced structs, array wrapper).
• Structural tests: every named type is defined exactly once and before use, for all test structs with and without the array record. IDL and AVSC declare the same record and enum names.
• New Device test struct: an enum inside a nested struct, and a record referenced three times.
• CLI tests: default output stays .avdl, --format avsc, case-insensitive format, invalid format rejected, file prefix, array record, valid JSON output.

Compatibility

No behavior change for existing users; idl remains the default and its output is unchanged.