Detect host id from function runtime environment

[andreasohlund]

This PR makes sure that the endpoint instances are uniquely identified as functions are scaled up or down.

The logic falls back to CONTAINER_NAME when running in Azure container apps and Environment.MachineName for local development.

NOTE: This is different from the existing worker package, which uses ENDPOINT_NAME. The reason for this was likely that ServicePulse at that time didn't have the "Do not track instances" feature yet, and we wanted a way to not continuously add instances in ServicePulse. This will be pointed out in the migration guide.

Design decisions

Align with endpoint instance concept

In order to align with the definition of endpoint instance, which states

Each endpoint has at least one endpoint instance. Additional endpoint instances may be added to scale-out the endpoint.

and also the documentation for heartbeats in ServicePulse, which states

Do Not Track Instances is the best setting for endpoints that are hosted in infrastructure with autoscalers. For example, Kubernetes, Azure Container Apps and AWS Elastic Container Service. Historical instances are not tracked, and only currently alive instances will be shown. The endpoint will be marked as unhealthy if there are no alive instances, and will be automatically removed after 6 hours of inactivity.

Each scaled-out function instance should be uniquely identified.

[danielmarbach]

Discussed it with @andreasohlund and after talking it through with @bording too we think going with calling AppContext directly

[andreasohlund]

@danielmarbach app switch now set in code, ready for review