When this project moved into GitHub, I spent some weeks changing terminology from slave to replica where possible. This included variable names, URL path, CLI params, and more. I kept slave terminology where:

• there was strong correlation to a MySQL command (e.g. stop slave)
• in some API responses where I didn't want to break usage.

This PR (WIP) pushes forward:

• In web interface, use replica
• In failure analysis, use Replica rather than Slave, e.g. DeadMasterAndSomeReplicas instead of DeadMasterAndSomeSlaves
  This breaks compatibility for people who rely on the specific analysis type (== text).

There are still a couple public API facing fields unchanged in this PR.