fix(retention): page through all expired records, remove 10k cap (#102)

[rajfirke]

What does this PR do?

RetentionEngine.find_expired() called trail.query(end=cutoff, limit=10000) — a single fixed-size query. Any trail with more than 10,000 expired records would silently drop the remainder: preview() under-counted, and execute() left those records undeleted while reporting success. For a compliance retention engine enforcing EU AI Act Art. 12 archival requirements, this is unacceptable.

Fix: Page through the full result set using offset:

batch = self._QUERY_BATCH_SIZE
expired: list[dict[str, Any]] = []
offset = 0
while True:
    page = self._trail.query(end=cutoff, limit=batch, offset=offset)
    expired.extend(page)
    if len(page) < batch:
        break
    offset += batch
return expired

_QUERY_BATCH_SIZE = 10000 is a class attribute instead of a hardcoded literal so tests can shrink it to 2 and exercise the multi-batch path without inserting 10k rows.

Two new regression tests verify that find_expired() and execute() correctly handle a trail whose expired record count exceeds one batch.

Checklist

• Tests added/updated for the change
• ruff check src/ tests/ passes
• ruff format --check src/ tests/ passes
• mypy src/provena/ passes
• pytest passes with no failures
• CHANGELOG.md updated (if user-facing change)

Related Issues

Closes #102

---

Note: PR #158 covered the same fix but was blocked waiting on offset support in ContextTrail.query() (which was already shipped in v1.1.0). Applying it directly here with a clean rebase on current main.