Also fixed a couple of bugs.

This way the forum and private group client do not need to keep track of
message timestamps themselves and do not need to interact with
post/message factories.

This is also lays the groundwork for #384

in group metadata to be able to speed up group listings.

Closes #584, #586, #585

This was happening when the remote response arrives before the local
response is made and thus the local response needs to be send with the
ACK following. The problem was that we ACK was sent before the response
which is not allowed and resulted in the session being aborted by the
introducee. This was happening, because recursion is hard ;)

The fix is only restarting another protocol engine to send the ACK
after the first run has been completed.

An integration test was added to prevent such regression in the future
and to test this code path.

Like other recursive operations on the dependency graph, this is
not done in a single transaction to prevent an attacker from creating
arbitrary large transactions.

So at startup, the `ValidationManager` finds and resumes any
unfinished operations, by looking for shared messages with unshared
dependencies.

This also adds unit tests to prevent regressions like this in the
future.

The docs say this can happen for a directory if there's an I/O error.

and use it for retreiving the status of an author faster.
Also add tests for both.

Before the introducee sends her ACK,
she derives a master key from the ephemeral shared secret as before.

Two nonces and a MAC key are then derived from the master key.
The local introducee signs one of the nonces and calculates a MAC
over her own identity public key, ephemeral public key,
transport properties and timestamp.
The local introducee includes the signature and MAC in her ACK.

On receiving the remote introducee's ACK,
the local introducee verifies the signature and MAC.
Should the verification fail, an ABORT is sent to the introducer and
the remote introducee that was added as inactive is deleted again.

The MAC and signature are not yet generated and verified.
This will happen in a later commit.

Comments and reblogs need to depend on the post they refer to.
Since message dependencies are limited to one group,
the post and also the comments need to be wrapped
when commented on or reblogged to another blog.

For this reason, in addition to comments, two new wrapping message types
are introduced. They retain all data of the original messages and allow
for reconstruction and signature verification.

This commit breaks backwards compatibility with old blog posts.
It removes the content type, title and parent ID from the post
message structure.