Composite readings

[eigenein]

Currently all the values are primitives. And for example a reading cannot be:

• an image annotated with a caption text
• an image annotated with security tags such as "has motion" or "has person"

Also, I still have to implement sensor titles somehow (#10).

A draft idea is to replace the primitive value: Value with a HTTP-headers-like set of values. For example, an annotated image could look like:

vec![
    (Header::Image, Value::ImageUrl("https://...")),
    (Header::Title, Value::Text("Hello")),
    (Header::SensorTitle, Value::Text("Backyard Cam")),
    (Header::HasMotion, Value::Boolean(true)),
]

And then renderer picks up everything it's able to render. Automation has to be adjusted too. Perhaps relates to reading transformations (#30).

Open questions:

• Does a reading need sort of "main" entry?
• Or does a reading still need to have a primitive and optional headers?
• How to prevent undesired unwanted headers? Say, two plain text titles are unwanted but one plain text with one in Markdown could still be useful.

[eigenein]

Potentially this could also help to get rid of the special "one-off" message type, leaving only read and write ones. Whether it's persistent or not could be put into metadata

[eigenein]

Trying to evolve this idea I decided that it's better to add strongly typed fields, either on message or reading level, depending on whether they should be stored per sensor or per reading