Add a generic attributes JSON column to machines for flexible per-machine extra data

[MMikePL]

Adds a generic attributes column to the machines table to store extra per-machine data — whether generic (e.g. last shutdown date) or machinery-specific (e.g. VNC connection details).

The goal is to avoid changing the DB structure for every new field. Any data that doesn't warrant a dedicated column goes into this JSON bag as a key/value pair, so upstream, custom and extension implementations can all add fields without changing DB structure. This keeps the schema stable across releases and avoids the compatibility problems that frequent structural changes cause for downstream forks.

For optimisation the column is stored as real jsonb on Postgres, with a fallback to plain JSON on other back-ends (SQLite/MySQL). Machine.attributes is a plain Python dict backed by a JSON/JSONB column. Changes are auto-tracked via SQLAlchemy's MutableDict — direct in-place mutation (machine.attributes["k"] = v) is fully supported.
The helpers are still provided as convenience wrappers: get_attribute(key) for safe None-guard reads, and set_attribute(key, value) which also lazy-inits the dict and removes the key when value=None.

The column name attributes is a proposal — it's a generic key/value bag, so properties, extra, params, or similar name would work equally well. Open to maintainer preference :)

[doomedraven]

this is interesting, i need the same for the tasks so i will review this soon and probaby extend, thank you for the idea!