This is a follow-up issue regarding a refactor for the function `_package_version_cmp()`.

As of now, there is only one item left to be addressed that is to improve the current function we have in the code `_package_version_cmp()`, which was improved in RHELC-297, but tkuratom@redhat.com made a few comments that we should address to improve the function itself.
 

 Acceptance Criteria

• Separate version parsing and version comparison
• The parsing function will take a version string in either ENVR or NEVR format.
• If epoch or arch does not exist in the input string, it will be set to None in the tuple.
• Parsing arch will be slightly tricky as I don't think there's a fool-proof way to separate an optional arch from release. Maybe looking for a list of allowed arch patterns is the best we can do. Something like: .*(x86_64|s390x|i.86)?$
• The version comparison function will take two version strings in either ENVR or NEVR format.
• The strings will be passed to the parsing functions to return the information in a normalized form.
• The names and arches will be compared.
• An error will be emitted if names do not match.
• An error will be emitted if arches do not match and both arches are not None.
• rpm.labelCompare() will then be used to compare the EVR for both packages and the results returned.
• Port existing use of pkghandler.compare_package_versions() to use the new comparison function.
• Remove pkghandler.compare_package_versions() and utils.string_to_version().

Notes

• Link for the thread discussion https://github.com/oamg/convert2rhel/pull/469#discussion_r873971400