FIX Surface AttackResultEntry.timestamp on hydrated AttackResult

[thirteeneight]

Description

Fixes #1651.

AttackResultEntry persists a non-nullable timestamp column, but get_attack_result() dropped it when rebuilding the domain AttackResult. The backend mapper's attack_result_to_summary then fell back to datetime.now(timezone.utc), so the UI showed today's date on every row no matter when it was actually persisted.

Three small, additive changes:

• Add an optional timestamp: Optional[datetime] = None field to the AttackResult dataclass.
• Pass timestamp=_ensure_utc(self.timestamp) in AttackResultEntry.get_attack_result().
• In attack_result_to_summary, prefer ar.timestamp over datetime.now() when metadata["created_at"] is absent. The metadata override path is preserved unchanged.

Backwards compatible: AttackResultEntries.timestamp has been non-nullable since the entry was introduced, so every existing row has a value to hydrate from. Callers constructing an AttackResult without the new field get None, which the mapper treats the same as missing metadata.

Filed as draft pending maintainer feedback on #1651.

Tests and Documentation

New unit tests:

• tests/unit/models/test_attack_result.py::TestAttackResultTimestamp — default is None, aware datetime is preserved, round-trip through AttackResultEntry, naive SQLite timestamps are normalized to UTC on hydration.
• tests/unit/backend/test_mappers.py::TestAttackResultToSummary — three new cases covering ar.timestamp preferred when metadata is absent, metadata still wins when both are present, fallback to datetime.now() when both are absent.

Verification:

• pytest tests/unit/ — 1339 passed, 0 regressions.
• mypy --strict on the three modified source files — clean.
• ruff format and ruff check — clean.
• No JupyText / docs changes: this is a bug fix and the only public addition is one optional dataclass field.

[thirteeneight]

@microsoft-github-policy-service agree

[behnam-o]

looks great to me, just some minor comments! thanks for finding this gap and addressing it!

[behnam-o]

I think if the entry already has a populated timestamp, we'd want to set it one memory model too, i.e:

    self.timestamp = entry.timestamp or datetime.now(tz=timezone.utc)

[thirteeneight]

Done, exact line you suggested. Round-trip's lossless now.

[behnam-o]

to be consistent with non-nullability of this in memory also, can we set it as

timestamp: datetime = field(default_factory=lambda: datetime.now(timezone.utc))

[thirteeneight]

In, ty. One ripple: get_attack_result() was passing _ensure_utc(self.timestamp) straight to the constructor, mypy started complaining once the field stoped being Optional. Mirrored your or datetime.now(tz=timezone.utc) on the hydrator side too. Tests updated for the new default and the entry-preserving path.

[behnam-o]

just an observation: I'm trying to find out where the metadata fields "create_at" is even set ... I can't seem to find any reference to it being set ...

I think we can fix that in a later change, @romanlutz if you know where this is populated, can you correct me please?

[thirteeneight]

Left this alone since you flagged for a later change. Happy to fold it in here if you'd rather not split it.

[thirteeneight]

Hey Behnam, thanks for the quick review! Both suggestions are in, dropped a reply on each comment with what changed. Let me know if it lines up with what you had in mind.