NEM-366 Ensure deterministic ordering on file export

[Mateus-Cordeiro]

This PR makes YAML exports deterministic by preserving model-defined field order when serializing dataclasses and Pydantic models.

PyYAML's default behaviour can produce noisy diffs due to unstable key ordering.

Changes

• Updated yaml.py's YAML representer to handle dataclasses and Pydantic models
• Generic objects (anything with dict) now serialize public attributes (non _ prefixed) with keys sorted alphabetically to avoid nodeterministic attribute order.

Limitations

Database samples can't be guaranteed to be deterministic. At the moment, a sampling query looks like this:

SELECT * FROM "{schema}"."{table}"

At the point in the code in which we obtain samples, we know the columns of the table, so an ORDER BY clause could be created. This, however, would not guarantee that the same rows would be returned.

[JulienArzul]

Database samples can't be guaranteed to be deterministic. At the moment, a sampling query looks like this:

SELECT * FROM "{schema}"."{table}"
At the point in the code in which we obtain samples, we know the columns of the table, so an ORDER BY clause could > be created. This, however, would not guarantee that the same rows would be returned.

Very good point, it would probably still be a good idea to use an ORDER BY on the samples (probably on the primary key?) so that at least two subsequent calls to build_context would return the same content if the table's data hasn't changed.
Without an ORDER BY, we're not even sure that we're getting back the same rows even if the table's data hasn't changed...