The information that we read from .changes and .deb files is also
available in projectb. It would probably be more efficient to just pull
it from there rather than doing bucketloads[tm] of I/O.

Control: retitle -1 release.debian.org: use projectb more in
debrelease.*

Thinking about it, there's more than strictly debrelease.proposed.

Things we currently use files for:

- determining which .changes files are in p-u-NEW
- reading Closes: from .changes (we already use the projectb record to
display Closes: for accepted packages)
- reading binary package lists from .changes
- extracting binary version information from .deb files
- extracting package stanzas from .deb files to feed to dose (in
debrelease.debcheck)