- Message stubs are now actually handled instead of being dropped on the
  floor, and are either converted into a warning or an actual message.
- As a side-effect of this work, missed calls are now properly
  announced! Yay!
- More work probably needs doing around notifying the user about
  ciphertext messages (in a non-spammy fashion)

- When messages are lacking in ciphertext (it needs to be reencoded for
  us), WA will send us a message stub (which we don't process, and need
  to).
- This then leads to us getting an empty message content, eating it
  to avoid saying something useless, storing the msgid, and then not
  sending the real message content later!
- This fixes that by checking we're actually sending something before
  storing.
  (It's only a short-term fix really though.)

- The version number was changed to 0.2.0, purely because rpm doesn't
  like '0.2.0-pre' and I got tired of switching it around.
- We now use the Unidecode library to produce ASCII representations of
  non-ASCII names, making life easier for people with umlauts and such
  in their contact names.

This can happen if someone tries to use revoked persistent session
credentials. Currently, we just hammer WA's servers trying to retry,
which is bad.

- We now only log WA connection state information at the DEBUG level,
  to try and avoid spamming the logs every time the websocket drops
  for whatever reason (which has become more frequent as of late)
- A new mechanism enables messages sent while offline to be queued
  and sent once the connection is reestablished, making small drops
  less of a pain.
  - This in fact re-uses the ack tracker code, so the user also gets
    alerted if the message doesn't send for a while.

- Contact nickname handling has been completely revamped, and now
  properly deals with things like nick collisions.
- In addition, the source of nickname information is now recorded,
  meaning that WhatsApp notify / contact name changes now propagate
  to the nickname changing, if the relevant option is enabled.

- ...but not much else.
- It works, though! Kind of.

- If `whatsapp.track_presence` is enabled, the bridge will request
  presence updates for users after you send them a message.
- The `GHOST PRESUB` command was also added, which manually attempts
  to subscribe to presence updates.

- We now use the fancy new improved WhatsApp connection logic
  in ww-rs-eta 0.5.0-pre1, which is nice.
- As a result, the handling of failed and superseded connections
  should be vastly improved; no more random QR code requests from
  lingering connections!
- In addition, the user now gets notified whenever they *remove*
  the connection from their mobile app (although it could also
  conceivably remove the state from the DB as well)
- History fetching was removed, as it never worked.
- The `ignore_other_libraries` config key now doesn't ignore
  whatsappweb_eta, because it's useful to have that in there
  for debugging as well.
- Some `dyn`s were dotted about the place to satisfy Rust 2018 lints.

- Messages now have timestamps stored in the database, so messages
  sent while the admin was away should be sent with timestamps now.
- There's an undocumented history fetch feature which doesn't quite
  work yet.

- Contacts' notify / name values were just being discarded (!) on
  contacts sync. This is now fixed.
- A new whatsapp.mark_read option can be used to control whether
  new incoming messages are automatically marked as read.
  This now defaults to off, as the old behaviour was somewhat
  confusing.

- Instead of contacts/recipients potentially being created in like
  three disparate places, there is now one and only one place
  where this happens: the object responsible for the protocol,
  i.e. the WhatsApp or modem manager.
  - This makes everything far simpler, and means we can actually
    keep track of things like contact names properly, instead of
    wondering which part of the bridge will end up making the
    contact in the end.
- All of the WhatsApp contacts we get on the initial burst, and
  any added thereafter, are now translated into recipients.
  - In addition, we keep track of their 'notify' property (i.e.
    the ~tildename for each contact, or the contact name from the
    user address book), and store that in the database.
  - In future, the bridge will perhaps let you autoname people,
    or display their notify in some meaningful way to allow you
    to differentiate people amongst the sea of random numbers...

- A new macro, sink_outbox!, was added to deduplicate the same
  "send messages into this futures sink from this outbox array"
  behaviour used in multiple places now.
- We now have the beginnings of an IRC server / s2c protocol impl!
  It handles client registration, and pretty much nothing else
  after that at this stage - but it's getting along!