fix/refactor strong group by tag processing, bucket fallbacks

[Dieterbe]

• clean '=' from 'group by' early on: turn the strong tags using '=' into a syntactic sugar that helps setting up the datastructure but is no longer visible afterwards, simplifying the default query structure and logic such as in target.get_graph_info().

  this fixes a bug where 'group by mountpoint=:dfs1' wouldn't do the right thing, because the bucket tag key didn't match the real tag key

• make sure buckets are unique (bugfix), and always have an explicit fallback. this makes the code more explicit and we can rely on it elsewhere
• add test cases for build_buckets and query parsing for a group by with a strong tag

@thepaul i want to merge this. but it breaks one of your test cases, i wasn't able to figure out yet how/why.

[Dieterbe]

never mind, the test was one my own, and i fixed it now. all tests pass!