PKCS7 parser tests: Replace fuzzer inputs by robust test cases

[gilles-peskine-arm]

The PKCS#7 unit tests include some test data that was found by fuzzing. As such it's somewhat “bizarre”, for example it uses an obscure hash (RIPEMD160), and it turns out that they won't reach the intended function once other validation bugs are fixed.

The goal of this issue is to construct robust test cases for the bugs fixed by 290f01b, e7f8c61 (if it's indeed a bug, if not explain why) and 4f01121. A robust test case should only have one thing wrong with the data, and is generally constructed by starting from a valid input and making one modification.

The test cases can be unit tests for auxiliary functions, they don't have to be for the main parser entry point.

Once we have robust test cases, we can remove the fuzzer inputs.