[阶段6] 同步 Garmin 全量数据：31 项日指标 + 奖励 + 个人纪录

原来每天只存 7 个指标，而 get_user_summary 一次就返回 60+ 字段，
另有睡眠分期、训练准备度、耐力分等独立端点从未被调用。

db.py:
- health_data 新增 31 列（距离/活动卡路里/基础代谢/爬楼/强度分钟/
  久坐时长/最高最低心率/最大压力/身体电量四项/血氧/呼吸/
  睡眠深浅REM清醒分期/睡眠血氧/睡眠呼吸/睡眠压力/训练准备度/
  VO2max/耐力分）
- 新增 badges 与 personal_records 两张表，均以 (user_id, garmin_id)
  为主键，重复同步更新而非累积
- 新增增量迁移: CREATE TABLE IF NOT EXISTS 对已存在的表不生效，
  新列必须显式 ALTER，否则生产库上永远不会出现。按列名比对后
  逐个补齐，SQLite 与 MariaDB 都幂等

services/garmin.py:
- _extract_daily 改为汇总 user_summary + sleep + hrv +
  training_readiness + training_status + endurance_score 五个端点
- 每个可选端点用 _safe 包裹：某项设备不记录时留 NULL，不影响当天其余数据
- 新增 sync_badges / sync_personal_records（账号级，每次同步取一次）

fix(garmin): 个人纪录整批写入失败
- Garmin 在同一份数据里混用 ISO 字符串和 Unix 毫秒时间戳，
  prStartTimeGmt 是 1570961412000，写进 DATETIME 列被 MariaDB
  以 1292 拒绝，导致 11 项个人纪录一条都没存进去
- 新增 _to_datetime 统一处理 ISO / 毫秒 / 秒三种形状，并优先取
  Garmin 自己提供的 *Formatted 字段

services/ai.py:
- 送给模型的 CSV 从 7 列扩到 23 列，纳入身体电量、血氧、呼吸、
  训练准备度、耐力分和睡眠分期

接口: GET /api/health/badges、/api/health/personal-records

tests (+13, 共 292):
- 徽章/纪录的往返、重复同步不累积、按用户隔离
- 两个用户可持有同一个 Garmin 徽章 id 而不冲突
- 时间戳三种形状的归一化及无效值不抛异常

NAS 实测: 7 天数据每天 31 项指标、65 个奖励、11 项个人纪录

Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>

backend/services/ai.py

@@ -361,13 +361,28 @@ def resolve_chain(preferred=None):
 
 
 # --- prompt construction ----------------------------------------------------
+# Kept deliberately short: every extra column multiplies by the number of
+# days sent, and the column names double as the vocabulary the model cites
+# back in `basedOn`.
 _CSV_COLUMNS = [
     ("date", "date"),
     ("steps", "steps"),
+    ("distanceMeters", "dist_m"),
     ("heartRate", "rest_hr"),
+    ("heartRateMax", "max_hr"),
     ("heartRateVariability", "hrv"),
     ("stress", "stress"),
+    ("stressMax", "stress_max"),
+    ("bodyBatteryHigh", "bb_high"),
+    ("bodyBatteryLow", "bb_low"),
+    ("spo2Avg", "spo2"),
+    ("respirationAvg", "resp"),
+    ("intensityMinutes", "intensity_min"),
     ("caloriesBurned", "kcal"),
+    ("activeCalories", "active_kcal"),
+    ("floorsAscended", "floors"),
+    ("trainingReadiness", "readiness"),
+    ("enduranceScore", "endurance"),
 ]
 
 
@@ -379,7 +394,10 @@ def build_prompt(summary, activities=None, day_budget=None):
     """
     day_budget = day_budget if day_budget is not None else default_day_budget()
     rows = summary[-day_budget:] if day_budget else summary
-    header = ",".join(label for _, label in _CSV_COLUMNS) + ",sleep_h,sleep_q"
+    header = (
+        ",".join(label for _, label in _CSV_COLUMNS)
+        + ",sleep_h,sleep_q,sleep_deep_s,sleep_rem_s,sleep_awake_s"
+    )
     lines = [header]
     for r in rows:
         cells = []
@@ -387,8 +405,9 @@ def build_prompt(summary, activities=None, day_budget=None):
             value = r.get(key)
             cells.append("" if value is None else str(value))
         sleep = r.get("sleep") or {}
-        cells.append("" if sleep.get("duration") is None else str(sleep["duration"]))
-        cells.append("" if sleep.get("quality") is None else str(sleep["quality"]))
+        for key in ("duration", "quality", "deepSeconds", "remSeconds", "awakeSeconds"):
+            value = sleep.get(key)
+            cells.append("" if value is None else str(value))
         lines.append(",".join(cells))
 
     sections = [