This adds new tool, apt2aptly, to mirror an existing apt repository in
aptly. To reuse obs2aptly's core logic, that was pulled out into a
shared crate, sync2aptly.

Signed-off-by: Ryan Gonzalez <ryan.gonzalez@collabora.com>

On full runs, this can drastically reduce the runtime (which also makes
testing significatly easier).

Signed-off-by: Ryan Gonzalez <ryan.gonzalez@collabora.com>

This requires refactoring in sync2aptly to generalize the location of
files, as well as backoff integration to retry on download / upload
failure.

Signed-off-by: Ryan Gonzalez <ryan.gonzalez@collabora.com>

This is how our actual layout on OBS is, so it needs to be matched here.

Signed-off-by: Ryan Gonzalez <ryan.gonzalez@collabora.com>

obs2aptly keeps its version computations, because sync2aptly is generic
over URLs and files, which makes moving it harder, and other origin
sources like apt2aptly can more easily obtain the versions anyway.
Unfortunately, the version can't just be pre-computed while gathering
the origin content; because that would result in a rather large amount
of version lookups in contexts where we currently don't need to. To
work around this, the version—and the function used to compute it—are
stored in a OnceCell + Box<dyn Fn> pair. *Technically* the caching isn't
entirely needed, but while reworking all of this code anyway, might as
well also avoid the reduntant version parsing.

Additionally, analyzing arch-independent packages switches from using
(source package, source version) pairs to (source package, package
version) pairs. This should have the equivalent effect of deduplicating
the packages for each architecture without relying on the source
version, which is hard to obtain from apt2aptly.

Signed-off-by: Ryan Gonzalez <ryan.gonzalez@collabora.com>

--apt-snapshot has been replaced with --apt-snapshots, which will point
to a file containing one snapshot to process per line. Already-existing
snapshots will now result in skipping the entire scan process, which
required some refactoring to get out a component list without also
doing a full scan.

Signed-off-by: Ryan Gonzalez <ryan.gonzalez@collabora.com>

This cleans up the code a bit and improves the parallel structure.

Signed-off-by: Ryan Gonzalez <ryan.gonzalez@collabora.com>

It's a bit clearer as to the actual purpose of the code.

Signed-off-by: Ryan Gonzalez <ryan.gonzalez@collabora.com>

This adds a new option, --publish-to, to publish to the given prefix,
as well as additional options to control signing and re-publishing
snapshots.

Signed-off-by: Ryan Gonzalez <ryan.gonzalez@collabora.com>

This is useful because the tool already knows the repo names to use.

Signed-off-by: Ryan Gonzalez <ryan.gonzalez@collabora.com>

Signed-off-by: Ryan Gonzalez <ryan.gonzalez@collabora.com>

aptly sorts the filenames before hashing them, but the .dsc files coming
from reprepro often aren't sorted.

Signed-off-by: Ryan Gonzalez <ryan.gonzalez@collabora.com>

We already do this in the normal OBS integration for performance
reasons.

Signed-off-by: Ryan Gonzalez <ryan.gonzalez@collabora.com>