**API**: Avoid compose ID parsing in productmd classes
So I've done some research over the last few days and come to
the conclusion that the compose ID parsing code I wrote for
fedfind is a lot more fragile and specific to Fedora's current
use cases than I thought it was. It certainly won't work right
for many RHEL cases. So to keep the productmd classes truly
generic, let's have `ProductmdComposeResult` take the full
composeinfo metadata dict as its input, not a compose ID as
before. This is, of course, an API change.
The Fedora classes still take a compose ID, but now find the
release from it and pass the full metadata dict to the parent
class. This is OK for current Fedora usage, and I intend to try
and robustify the compose ID parsing in fedfind as much as I
can with my new knowledge of productmd's behaviour, and restrict
it to just discovering info we need to find the release from the
CID.
Signed-off-by: Adam Williamson <awilliam@redhat.com>