[skipruntime] Allow to wait for external resources

[skiplabsdaniel]

No description provided.

[bennostein]

First batch of comments. Mostly little things, but a handful of questions/comments that could use responses.

I don't see any tests about the main bit of behaviour that I thought this was designed to address: the initial GET request always returning an empty collection when it depends on something external. Can you explain what observable changes this PR makes from the user's perspective? What guidance should we give on how to use this / handle loading of external resources?

I'll follow up with a commit to clean up the doc comments' English, but appreciate all of those!

[skiplabsdaniel]

The general principle is to add a collection of statuses to each resource, with each external service involved in the resource adding its status to this table, along with “creation” and “modification” times.
When a request is made, a “request” (for want of a better word) is created to mark that a request is being made at that time.
The various statuses are aggregated relative to the time of this request (all this to avoid unfortunate time sequences that could lead to infinite waiting).
A request can be made with a “Checker” (too, for want of a better word).
If a “Checker” is provided, if will be called at status change for the requester to check and do what is necessary.
If there is no “Checker” provided and there is a dependency being loaded a “request” identifier is returned to allow subsequent status request relative to that request.
This is the general case for having a notion of status for the resource throughout its life in the service.
There are still some unanswered questions:

• What does this mean for a Skip service chain (status propagation, error handling, etc.)?