`apps:logs` is not a real log solution

[stuartpb]

Plushu deserves a proper, first-level logs plugin and command.

What's wrong with docker logs -t $(cat $app_dir/proc/web.cid)? Compare it to heroku logs $app. To sum it up, the solution here only gives logs from the running instance, and doesn't include any other logs- like, say, logs from a crash. (Oh yeah- we're not restarting the app when it crashes. That's another issue.) It doesn't include build/release output. It doesn't include request logs. It doesn't include timestamps. Frankly, it sucks.

This might be doable within the scope of continuing to use progrium/buildstep (at least for building), or it might involve overhauling the inside of the container too.

Right now, I'm unconcerned enough with logging to let this slide.

[stuartpb]

Some relevant literature: https://blog.logentries.com/2014/03/the-state-of-logging-on-docker/

[stuartpb]

You know, if sort can do timestamps, this miiight be doable with what we have now, although there are still questions:

• How should lifecycle steps be logged? (Right now they barely even output anything. Will something (receive) log their stdout? Will the plugins need to be modified to call out to a logging hook?)
• What handles restarting crashed containers, getting/saving their output, etc?

[stuartpb]

Further discussion is being held in plushu/plushu-logs#1. (Was apps:logs ever even implemented?)