Handle duplicate packages in upstream repos
Some repositories have some packages listed in the metadata twice. This shouldn't happen (but it does). createrepo_c deduplicates by virtue of parsing everything into a dict keyed by pkgid, but the iterative parser does not. This eventually results in a mismatch once the iterative parser comes across a package that the createrepo_c primary parser already handled. So we keep a list of pkgid's we've already written out in order to skip them once the iterative parser hits them a 2nd (or 3rd, etc.) time.