feat: Phase 4 — Experiment 自动归档 + 异步 Job（状态机 / SSE / 一键复跑）

- 数据表：job / experiment（spec/result JSON 存档、code_version），Alembic 迁移 53113c80257f
- Job：queued→running→(success|failed) 状态机，BackgroundTasks 本地执行 + 失败兜底标记；结果与 Experiment 关联
- Experiment：每次研究成功自动归档（含 git commit 与收益摘要），支持一键复跑（同 spec 重建 Job）
- API：POST /api/jobs、GET /api/jobs/{id}（内嵌结果）、SSE /api/jobs/{id}/events、/api/experiments 列表/详情/rerun
- 前端：新增「实验」页（列表 / 详情 / 复跑 + Job 轮询）；导航更新
- 端到端验证：真实 20 股 job 提交→后台执行→success→EXP 归档（-12.41%）；executor 成功/失败路径单测
- 测试 73 passed（新增 5 项 Job/Experiment）/ ruff clean / 前端 tsc + build 通过

backend/app/infrastructure/persistence/sqlalchemy/repositories/jobs_impl.py

@@ -0,0 +1,63 @@
+"""Job / Experiment 的 SQLAlchemy 实现（Phase 4）。"""
+
+from __future__ import annotations
+
+from sqlalchemy import select
+from sqlalchemy.orm import Session
+
+from app.domain.entities.research import ExperimentRecord, JobRecord
+from app.infrastructure.persistence.sqlalchemy.models.jobs import ExperimentModel, JobModel
+
+
+def _job_to_model(job: JobRecord) -> JobModel:
+    return JobModel(**job.model_dump())
+
+
+class SqlAlchemyJobRepository:
+    def __init__(self, session: Session) -> None:
+        self._session = session
+
+    def create(self, job: JobRecord) -> JobRecord:
+        self._session.add(_job_to_model(job))
+        self._session.flush()
+        return job
+
+    def get(self, job_id: str) -> JobRecord | None:
+        row = self._session.get(JobModel, job_id)
+        return JobRecord.model_validate(row, from_attributes=True) if row else None
+
+    def update(self, job: JobRecord) -> None:
+        row = self._session.get(JobModel, job.id)
+        if row is None:
+            raise KeyError(f"job {job.id} 不存在")
+        for k, v in job.model_dump().items():
+            setattr(row, k, v)
+
+    def list_recent(self, kind: str | None = None, limit: int = 20) -> list[JobRecord]:
+        stmt = select(JobModel).order_by(JobModel.created_at.desc()).limit(limit)
+        if kind:
+            stmt = stmt.where(JobModel.kind == kind)
+        return [
+            JobRecord.model_validate(r, from_attributes=True)
+            for r in self._session.scalars(stmt).all()
+        ]
+
+
+class SqlAlchemyExperimentRepository:
+    def __init__(self, session: Session) -> None:
+        self._session = session
+
+    def save(self, experiment: ExperimentRecord) -> ExperimentRecord:
+        self._session.add(ExperimentModel(**experiment.model_dump()))
+        self._session.flush()
+        return experiment
+
+    def get(self, experiment_id: str) -> ExperimentRecord | None:
+        row = self._session.get(ExperimentModel, experiment_id)
+        return ExperimentRecord.model_validate(row, from_attributes=True) if row else None
+
+    def list_recent(self, limit: int = 50) -> list[ExperimentRecord]:
+        rows = self._session.scalars(
+            select(ExperimentModel).order_by(ExperimentModel.created_at.desc()).limit(limit)
+        ).all()
+        return [ExperimentRecord.model_validate(r, from_attributes=True) for r in rows]