fuzzing: fix serialization of strings containing null bytes

[dguido]

Summary

• Fix serialization of strings containing null bytes (\x00) and other non-printable control characters
• Strings that cannot be reliably represented in JSON are now hex-encoded with a hex: prefix
• Add comprehensive unit tests for round-trip serialization of strings with null bytes

Problem

When serializing ABI values to JSON for corpus storage, strings containing null bytes were being incorrectly saved as empty strings. This is because JSON cannot properly represent arbitrary binary data in strings - null bytes and other control characters can be lost during the []byte -> string conversion or JSON marshaling.

Solution

The fix:

1. Add stringNeedsHexEncoding() helper to detect strings that contain null bytes, invalid UTF-8 sequences, or control characters (except common whitespace like \t, \n, \r)
2. In encodeJSONArgument(): Encode such strings as hex:<hex-encoded-bytes>
3. In decodeJSONArgument(): Decode strings with the hex: prefix back to their original bytes

This approach:

• Maintains backward compatibility with existing corpus files containing normal strings
• Only applies hex encoding when necessary
• Follows the same pattern used for BytesTy serialization

Test plan

• Add TestStringWithNullBytes - verifies round-trip serialization for various null byte scenarios
• Add TestStringNeedsHexEncoding - verifies the helper function correctly identifies strings needing encoding
• Run existing tests to ensure no regressions: go test -v ./fuzzing/valuegeneration/...
• Run linter: golangci-lint run --timeout 5m

Fixes #279

---

Generated with Claude Code