no unbound recursion in publish()

unbound recursion approach can blow up call stack,
and - I think - allocate memory unboundedly as well.

We can simply loop until err != nil

I didn't actually test this live, though tests succeed

Lots of progress on InfluxDB Query Editor, show tag keys, values and measurement dropdowns now take into account already selected tags or measurements, so you can start with tag keys and the select measurement, #1525

fix mac.md for incorrect href

Minor work on InfluxDB 0.9 query editor, #1525, hit a roadblock with the new InfluxDB query language, does not support tag and tag values discovery queries with filters, not sure if I need to move to SHOW SERIES style queries

Addresses #1853, redact session provider secrets

In cases where a database is used for session storage, redact the
session_provider config value. I assumed "@" as the marker for a
database vs file/memory.

Backend configuration: When unknown key was detected the log error was not visible in log file due to logging was initialized after user config file is loaded, logging is now initialized twice, once based on defaults and command line arguments, then then again after user config file, Fixes #1992

Fix minor grammar issue in docs index page

Major refactorings around searching, moved to seperate package, trying to move stuff out of models package, extend search support searching different types of entities and different types of dashboards, #960

Rewriting anb abstracting how dashboards are loaded, unifying db, json files, and script dashboards, #960