Add property-based and fuzz testing for codecs and stream resynchronization

[wuisabel-gif]

Summary

Add property-based and fuzz tests for the Python, Rust, and C codec surfaces, with particular focus on malformed frames and stream resynchronization.

Motivation

The existing golden vectors verify known examples, but malformed inputs and arbitrary byte streams have a much larger state space. Property-based and fuzz testing can uncover length-field errors, integer boundary problems, parser hangs, inconsistent rejection behavior, and cases where corruption causes unnecessary loss of later valid frames.

Proposed properties

• Decoding arbitrary bytes never crashes, hangs, or performs unbounded allocation.
• Valid packets survive decode(encode(packet)) round trips.
• Frames with invalid CRCs are never accepted.
• Python, C, and Rust agree on valid and invalid corpus entries.
• A stream receiver eventually recovers after arbitrary garbage followed by a valid frame.
• Corrupt input cannot prevent later complete valid frames from being recovered indefinitely.
• Boundary values for packet size, counts, timestamps, and sequence wraparound behave consistently.

Proposed scope

• Use Hypothesis for the Python codec and stream receiver.
• Use proptest or cargo-fuzz for Rust.
• Add a C fuzz target compatible with libFuzzer or AFL++.
• Maintain a small shared regression corpus for interesting failures.
• Run bounded property tests in normal CI and deeper fuzzing in a scheduled workflow.

Acceptance criteria

• Python property tests cover valid packet generation and arbitrary byte streams.
• At least one native fuzz target is available for the C or Rust decoder.
• CI runs a deterministic, time-bounded test configuration.
• Discovered failures can be saved as shared regression corpus entries.
• Documentation explains how to run longer fuzz sessions locally.
• Sanitizers are enabled for native fuzz targets where supported.