explore alternative impl for cycle handling

[nikomatsakis]

In #147, I noted an alternative impl strategy for cycle handling that I wanted to explore:

Anyway, I also started a branch that was implemented a bit differently. What I was trying to do was to have each active query have an associated Arc with a pointer to the next active query on the stack (i.e., what it is blocked on). This way you could follow a chain of queries forward across threads pretty cleanly and lazily. I seem to recall that the code in this PR was a bit more complex -- but then I don't think I've looked at it since you rebased.

Filing this issue so I don't forget. Still hoping to explore this, but I'd also be happy to work with someone else on it.

[carljm]

@nikomatsakis Is this still relevant after #603 ?

[Veykril]

I believe we can close this now, the motivating comment is no longer relevant either