Frame hook rework for frame boundary synchronization

[stv0g]

As discussed with @pjungkamp and @fwege on Tuesday.

[stv0g]

This PR also adds a CI job to validate the OpenAPI spec in VILLASnode rather than the documentation repo.

The generated OpenAPI documentation can be viewed here: https://acs.pages.rwth-aachen.de/-/public/villas/node/-/jobs/6719350/artifacts/openapi.html#tag/config

[n-eiling]

Looks like a good change. Some small notes below.

[stv0g]

Okay, I spent far to much time on this now :D

I did some more refactoring:

• Replace enum Trigger with a std::variant
• Make parse_duration() aware over overflows, underflows and unsufficient precission (e.g. parsing "1us" into std::chrono::millisecond.
• Minor tweaks to code-style and readability.

[pjungkamp]

Just some small nitpicks for now.

[n-eiling]

LGTM
I'm getting a bit of a bad feeling that you wrap your custom smart pointer of Sample with a unique_ptr, but I guess this makes a future transition away from the manual reference counting easier.

[pjungkamp]

I'm also not too happy with using a unique_ptr for shared ownership, but it is reasonable solution to help with Sample leaks.

But there should be no reason to use a Sample::Ptr/shared_ptr<Sample> ever. Please remove that alias.

[stv0g]

Please remove that alias.

Done.