Add functionality for working with RECORD files

[SpecLad]

This somewhat belatedly resolves #708.

This is not a complete patch; the functionality is implemented, but there are no tests or documentation. Before I add those, I'd like to get some initial feedback on the API (and, of course, on whether you'd want to accept something like this at all).

To summarize, the API works as follows. You parse a RECORD file like this:

record_set = parse_record_csv(contents)

for record in record_set:
    print(record.path)
    if record.hash is not None:
        print(record.hash.algoritm, record.hash.digest)
    if record.size is not None:
        print(record.size)

And you render one like this:

builder = RecordSetBuilder()

builder.add(Record('a/b/c.py', hash=RecordHash("sha256", b'\x01\x02...'), size=123))
...

record_set = builder.build()
contents = record_set.to_csv()