Solve the DQlite issues

[sparshev]

In general DQLite was a good start for the project, but it seems we need to move on. There is a number of issues with using DQLite:

• Raft-based database growth
• Auth in the cluster is questionable
• Support of MacOS and Windows not to mention the other systems looks like a pain
• Performance is far from good
• Splitbrain concern is unclear
• C-nature of the dependencies makes it hard for cross-platform build process

So I think it will be better to switch to using sqlite for the node with a custom cluster communication protocol. The info distribution could be done via websocket and the nodes are establishing those connections to a limited number of neighbors to properly auth in the cluster via certificate and sign their requests to verify based on pubkey.

There is still a question how vote process will look like, and how the future web interface will collect the cluster information, but probably it's doable.

Expected Behaviour

DQLite need to be removed and replaced by sqlite and custom websocket-based cluster communication protocol.

Actual Behaviour

For now the Fish uses a number of C-based dependencies including DQLite

[sparshev]

Ok this ticket is finally closed, the only issue was not resolved is cluster sync - but since it not worked well in dqlite than it's a separated one #30.