History handling

[arnsholt]

In the IPython model, it is the kernel's responsibility to keep track of code history and their associated outputs. We currently don't. Some of the machinery is in place (most notably @!history, although it needs to store more stuff to handle every case expected by frontends), but full handling isn't implemented yet.

Also, it's not entirely clear to me whether we have to keep track of the entire command history all the time, or if we can prune it at some point, nor what exactly the output tracking entails: should we track all outputs, including errors and such, or only successful executions?