Ensure nodes can reconnect after dropping out

[whitfin]

This is based around #12 and corrects several issues in remote caches, namely;

• fixes an issue where caches were not linked to supervisors correctly
• fixes an issue where nodes are not brought into the cluster
• fixes an issue where tables are not replicated across the cluster
• adds a new add_node/2 function to add a node to a cache at runtime
• adds a new start/1 function to start a cache without proc linking
• adds new integration tests based around adding remote nodes

[coveralls]

Coverage remained the same at 100.0% when pulling 84e7051 on issue-12 into 403e1fa on master.

[whitfin]

I'm gonna let this soak for a little while and come back to it fresh before reviewing. The test setup is kinda flaky but seems to work 9 times out of 10 so is sufficient (at least it's good enough for me currently).

[coveralls]

Coverage remained the same at 100.0% when pulling e5a8ce6 on issue-12 into 403e1fa on master.

[coveralls]

Coverage remained the same at 100.0% when pulling 19b02c1 on issue-12 into 403e1fa on master.

[coveralls]

Coverage remained the same at 100.0% when pulling 24d875d on issue-12 into 403e1fa on master.

[coveralls]

Coverage remained the same at 100.0% when pulling acfdea4 on issue-12 into 403e1fa on master.

[coveralls]

Coverage remained the same at 100.0% when pulling 15b21fb on issue-12 into 403e1fa on master.