Investigate txtar failures and why the fuzzer is not working on Windows

[bufdev]

When running the fuzzer tests on Windows, they are failing because the txtar package does not seem to parsing the corpus files into the Archive correctly (the contents of the corpus files are being parsed as Comments).

Putting in a dummy comment somewhat addresses the issue (a different error is raised in the test clean up). For now, we have moved the fuzzer tests to unix only, but this is worth investigating in the future.