fuzz: add sv2_message_roundtrip fuzz target

[xyephy]

Add a fuzz target that tests serialization/deserialization roundtrip for all SV2 message types:

• Sv2SetupConnectionMsg
• Sv2NewTemplateMsg
• Sv2SetNewPrevHashMsg
• Sv2SubmitSolutionMsg
• Sv2CoinbaseOutputConstraintsMsg
• Sv2RequestTransactionDataMsg

This complements the existing sv2_noise_cipher_roundtrip fuzzer by covering the message parsing layer.

Edge cases to cover:

• Maximum field lengths (255-byte strings, 65K merkle paths)
• Boundary values (0, MAX_INT)
• Malformed/truncated messages

[Sjors]

That seems like a good idea. Do you know if the SRI codebase has any test vectors that we could use a starting place for a corpus?

[xyephy]

That seems like a good idea. Do you know if the SRI codebase has any test vectors that we could use a starting place for a corpus?

Yes, there's a dedicated fuzzing corpus repository: https://github.com/stratum-mining/stratum-fuzzing-corpus .
It contains seed inputs for SetupConnection, SetupConnectionError, SetupConnectionSuccess, SV2 frames, and data
types that we can use as a starting corpus.

[Sjors]

It would be nice if we can use that.

[xyephy]

It would be nice if we can use that.

Question: Is it a good idea to update the existing sv2-tp-qa-assets( corpus files) by copying SRI corpus (renamed to match target names) and add hand-crafted seeds for TP-specific messages?

[Sjors]

@xyephy I doubt that would work. The effect of a corpus file depends on the exact fuzzing harness code that runs it.