Easy API for having context span threads

[itamarst]

Fixes #225.

[exarkun]

Lines 1424 and 1425 in a helper function that this test method and test_no_context can both call?

❓

[itamarst]

I think with only two callers and two lines it doesn't end up being shorter. But yeah, if there were more of them.

[exarkun]

Thanks. This is a nice addition.

Relatedly, this past week I've been noticing that even some single-threaded code manages to lose the parent/child association - if it's asynchronous. Lots of code uses Deferred without being Eliot-aware at all. Any time execution has to jump to a callback on a Deferred, the Eliot context gets lost. That's a bit sad.

It seems like this preserve_context is part of a solution for that problem. It doesn't have to be used across threads, after all. Though maybe "remote_task" isn't the best way to represent "asynchronous mumble mumble" (it's not how DeferredContext glues things together, after all). But it doesn't help with the "this code uses Deferred and doesn't know about Eliot problem" though ... so ... I dunno.

I guess thread-using code may possibly suffer an equivalent problem. This helper doesn't help if you don't use it and you may want to call existing code that doesn't know about Eliot. Not sure what to do about that, though.

Anyhow, this looks like a useful helper for the case where Eliot-aware code does control the dispatch-to-thread piece, so, great.

Just a few minor points inline. Address to your satisfaction and then merge.

[itamarst]

Yeah, making stuff Just Work is hard with async and threaded code. A sub-class of Deferred might work in some cases, but sooner or later you'd hit a Deferred someone else created.