Change builder setup in package datasets for efficient reuse

[mih]

Right now each package dataset has exactly one builder. This makes it necessary to have additional package datasets for when feeding multiple distributions.

I believe this makes things needlessly complex. When backporting sometimes the source packages only change in terms of their version, sometimes not at all. The orig sources never change. So making it possible to use a single package dataset for more than one distribution would reduce overall waste and maintenance effort.

A number of changes are required

• deb-build-package needs to become configurable re builder choice (or become more clever by inspecting the a package (changes file) for the target distribution.
• deb-update-reprepro-repository cannot simply rely on any updates from a package dataset in a distribution dataset to actually be exclusive to this distribution (again needs to become more clever)