decide what to do with take_over

[chadwhitacre]

@zwn has identified take_over as a big source of bugs in our codebase. The purpose of that method is to allow users to merge one Gittip account into another. There are some UI problems that mean users end up in this situation more often than they should (#34), but we'll always have this use case no matter what.

Options:

1. kill take_over—Users can't merge accounts. If I have two accounts and I want to close one in favor of the other, I have to give up money, because if I have tips attached to one account, I have no way to "manually" reattach those tips to another account in the way I can "manually" move my statement, connected accounts, etc.
2. minimize take_over—Only support moving tips from one account to another but nothing else.
3. fix take_over—Recommit to a robust solution for merging accounts. Users get a UI to merge all aspects of an account into another (cf. merge ui should give you the option to make the other account primary #675, merging accounts doesn't move Balanced account #818, merging accounts should merge statements #834).

Compare this with GitHub and stars on repos: the stakes are higher for us because we're talking actual money, not just stars. I think we need at least (2), and I think (3) is better for users.

[chadwhitacre]

Labeled as DevX ★ because this ticket is about deciding our approach. We'll reticket a specific course of action once we decide on one.

[chadwhitacre]

Downgraded to DevX because I'm not sure this is actually super-high priority.

[zbynekwinkler]

We are going to need parts of take_over anyway for other, so far unsupported, functionality. I'd like to refactor take_over into separate testable functions usable for example for account suspension and minimize its use at the same time. I think having a lots of users ending up in take_over means that we are fixing a problem after it has happened instead of preventing it in the first place. One of the things we certainly need to do is separating "opt in" from recurent "sign in" to stop inadvertent opt ins that need to be fixed with take_over.

[Changaco]

Closing in favor of #2630.