Handle function instance creation failure

[ccicconetti]

When an edgeless_node starts a function in

edgeless_node::base_runtime::runtime::RuntimeTask<FunctionInstanceType>::start_function()

it optimistically assumes that the creation of the function instance will not fail, in fact, function_instance_runner::FunctionInstanceRunner::new() does not return a Result but an object.

However, with a container run-time the creation may fail for a number of good reasons (e.g., Docker daemon not responding, image not available locally), therefore the failure should be handled to avoid the current situation:

• function start
• instance creation fails -> the function is not initialized and an exit event is generated
• however, the node marks the request as well-accepted
• the orchestrator and controller receive no indication that the function instance is not running properly

[ccicconetti]

Furthermore, the exit of a function (and a resource) should be handled.

Currently, such an event goes unnoticed and the ε-ORC does not know that the function is not running properly (hence can take no corrective actions).