Improve tail reconnection logic

[esatterwhite]

The error handling logic for the websocket can be a bit wonky. If the server restarts or goes away - which happens usually because of kubernetes rollouts.

error: Unexpected EOF reconnecting
ERROR lost connection
ERROR [0]:

Users report that it will reconnect and continue just fine. But will occasionally will get duplicates lines when this happens.
Its likely that the server went down before that batch of lines was acknowledged and its being retried.

• Make the error messaging a bit more useful or indicate we are reconnection
• add an exponential backoff to reconnections.
• Add a maximum number of reconnections and exit when reached.