feat(retention): add export/clean/rehydrate endpoints for task content

Adds an operational surface for bounded retention of task chat content
in shared infrastructure. Callers can snapshot a task's content,
delete it from the shared stores, and later restore it byte-identically
from the snapshot — preserving message IDs and timestamps so tool-call
and reasoning references remain valid.

Three new endpoints under /tasks/{task_id}:
- GET /export — returns a self-contained snapshot (messages + task_states)
- POST /clean — deletes content across Mongo messages, Mongo task_states,
  Postgres events; resets agent_task_tracker cursors; sets tasks.cleaned_at
- POST /rehydrate — restores content from a snapshot, clears cleaned_at

Domain layer lives in TaskRetentionService so the eventual scheduled
sweep workflow and the HTTP endpoints share the same code path.

Cleanup uses a "Mongo deletes first, Postgres marker last" order so
retries after partial failure converge correctly. The active-task,
idle-threshold, and unprocessed-events guards refuse cleanup when the
task isn't safe to drop.

Schema:
- New nullable tasks.cleaned_at column (TIMESTAMPTZ, metadata-only ALTER)
- No new audit table — cleanup operations emit structured log lines

Other changes:
- adapter_mongodb.batch_create now translates pymongo BulkWriteError
  with all-duplicate-key sub-errors into DuplicateItemError (HTTP 400)
  instead of letting it surface as ServiceError (HTTP 500)
- New EventRepository.delete_by_task_id and
  AgentTaskTrackerRepository.reset_cursors_for_task methods

Tests: 13 integration tests covering happy paths, all precondition
guards, and the byte-identical export → clean → rehydrate round-trip.

Author: smoreinis

agentex/database/migrations/alembic/versions/2026_05_19_1929_adding_task_cleaned_at_6c942325c828.py

@@ -0,0 +1,26 @@
+"""adding task cleaned at
+
+Revision ID: 6c942325c828
+Revises: a9959ebcbe98
+Create Date: 2026-05-19 19:29:34.858692
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = '6c942325c828'
+down_revision: Union[str, None] = 'a9959ebcbe98'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.add_column('tasks', sa.Column('cleaned_at', sa.DateTime(timezone=True), nullable=True))
+
+
+def downgrade() -> None:
+    op.drop_column('tasks', 'cleaned_at')

agentex/database/migrations/migration_history.txt

@@ -1,4 +1,6 @@
-9ff3ee32c81b -> e9c4ff9e6542 (head), add_tasks_metadata_gin_index
+a9959ebcbe98 -> 6c942325c828 (head), adding task cleaned at
+e9c4ff9e6542 -> a9959ebcbe98, finalize_spans_task_id
+9ff3ee32c81b -> e9c4ff9e6542, add_tasks_metadata_gin_index
 57c5ed4f59ae -> 9ff3ee32c81b, uppercase deployment status enum labels
 4a9b7787ccd7 -> 57c5ed4f59ae, add_task_id_to_spans
 d1a6cde41b3f -> 4a9b7787ccd7, deployments