Fix context chain closing

Fix context chain closing

- Close and create EOS registration correctly when closing context
chain holder and creating context chain holder.
- Cancel running scheduled task on context chain holder termination
- Always remove device from DS when we was not able to propagate roles,
not only when we have this device already in context chain map
- Add toleration for adding context chains to queue
- Wait for registration of RPC services before fully granting mastership
for context chain
- Close transaction chain manager on device context close
- Remove unnecessary auxiliary connection gathering in device context
(because this was moved to context chain)
- Fix closing of TxChain when DeviceContext is closed when activating tx
chain manager but not enabling submit transaction (do not try to submit
data)
- Unregister RPC registrations on RpcContext close
- Do not close ClusterSingletonServiceRegistration during
ClusterSingletonService initialization, to prevent freeze of
ClusterSingletonServiceGroup, send it to separate thread
- Prevent propagating of connection closed event twice (add checks for
RIP connection state)
- Fix ignored tests
- Remove getter for table statistics boolean from DeviceContext and just
send it to DeviceInitializer
- Close all RPC contexts when RPC manager is closed
- Do not reschedule statistics one more time when context chain is
closed

Resolves: bug 8668

Change-Id: Ifa8c492486c17a73eccd760fc610a435ef6a42ed
Signed-off-by: Tomas Slusny <tomas.slusny@pantheon.tech>