This has two benefits:

1. Using channels makes it easier to time out while waiting
2. Allows us to clean up goroutines that were waiting if we timeout the
   request

Including:
- Handle timeout=0 correctly
- Always return immediately on initial sync
- Handle spurious wake ups from the notifier

* Turn on 'unparam' linting

* Increase deadline

* Fix mediaapi integeration test

* Remove unused struct field

* Ignore unused test data

* Remove unused variables

* Remove deadcode

* Fix up vetshadow warnings

* Convert to using gometalinter

* Update travis

* Use vendored versions of gometalinter

* Make gometalinter install its stuff

* Vendor misspell

* Use BuildEvent method on room join

* Fix building the list of room members in the sync notifier

* Fix building the list of room members in the sync notifier

* Rephrase comment

* Clean roomserver consumer

* Make account data sync incremental

* Use a different name for the sync AD table

* Improved error logging

* Created missing topic in tests

* Add client API topic to tests

* Add client API topic to common

* Move data batch retrieval

* Add database index for data retrieval

* Fix typo in table name

* Fix indentation

* Use utility methods from gomatrixserverlib, rather than reimplementing them

* Return string rather than pointer to string

* Update gomatrixserverlib

- `s/Server/OutputRoomEvent/` in `consumers` to accurately reflect what is being consumed.
- `s/set/userIDSet/` in `notifier.go` for clarity.
- Removed lying comments.

* Split out notifying /sync requests and calculating sync responses

The logic for notifying /sync requests is about to get really
complicated as we optimise when to wake up requests, so split
out that code into a separate struct to isolate it and make
it easier to unit test.