Figure out design for handling errors

[istathar]

Basic functionality for tracer is working!

I think what's missing, however, is a notion of what to do in the case of encountering an error. This seems to have two parts

1. detecting an error has occurred in a child process
2. causing a root span to be emitted anyway

Detecting errors

The first one smacks a bit of being a "long tail" problem. GitHub Actions, for example, attempt to highlight lines with warning and error which is cool! though I note that the build traces out of GHC didn't seem to trigger their heuristic.

A first approximation is exit code, of course, and filing an "error" span on a non-zero exit code would be a good step in the right direction.

Ensuring a root span

The second of these is tricky. While we have finally and onException type constructs in real programming languages, in a YAML driven environment having a program that runs before exit no matter what is trickier.

I don't know of any good way to deal with this, other than saying "ensure you run this command at the end no matter what" is the requirement, but it's weak.