feat(session,runtime): 多角色并存—按会话选角（迁移0011 + 按回合解析 + agent池 + 提取器修正）

按 DevDocs/design/22 批次 1 后端（§2/§3/§4.1-4.5/§4.7/§5 验收 1-6）。
- 迁移 0011：chat_sessions.persona_id + chat_messages.persona_id（均 DEFAULT ''，无回填 → 升级零风险）；同步包内幂等 DDL；**真实库升级验证**（拷贝 data/cyrene.db 到临时目录跑 MigrateUp：既有 1 会话/109 消息 persona 仍为 ''，原库未被触碰）
- history：SessionMeta/Message 加 PersonaID + SetSessionPersona；ListSessions/List/scan/Append 带上
- runtime：SessionEnv.PersonaID + loop.personaFor(ctx) 替换固定 l.personaID（memory/request/turn）+ **agent 池**（Options.Agents，按角色取 instructions，未命中回落）+ **提取器不再把 persona 烤死在构造期**（多角色下记忆不再永远记在 cyrene 名下）+ **thinking 旋钮改为按会话角色解析**（否则角色 Y 会用角色 X 的预算/展示）
- gateway：会话 DTO/POST/PATCH 支持 personaId；ws/rest/voice 的 OwnerEnv 硬编改为**按会话解析角色**（actor 元数据语义不变）；admin memories 搜索去掉硬编 cyrene
- Docs/api-client v0.42；**两处有意的行为收紧**已写明：POST /sessions body 非法 JSON → 400（原先完全忽略 body；空/null/{} 仍 201）；POST /admin/memories/search 省略 personaId 且无默认角色 → 400（原先默认查 cyrene）
- 测试：design/22 §5 验收 1-6（记忆隔离/提示词首条 system/情绪按角色/提取 persona 归属/迁移兼容）+ 进程内 e2e（断言 adapter 收到的 model.Request.Messages[0] 即该角色的 systemPrompt）
- **已知未做（留批次 2 / 待确认）**：① IM 与渠道级未做 → **§7-3「owner 私聊镜像随主会话角色」尚未生效**；② 请求级覆盖未实现（本批只做会话级）；③ **§7-5 的「上下文发送角色标记」只完成存储+DTO 半边**——消息 persona 已落库且 DTO 已带，但模型上下文不从 chat_messages 重建，「第 N 条是谁说的」尚未进上下文（需确认是否属批次 2）；④「切角色重建摘要」当前是空操作（压缩未接线）；⑤ thinking 旋钮仅在流式路径生效（非流式 agent.RunStep 不带，A 批次既有性质）
- 验收：go build/vet/test 全绿 + linux/amd64 与 darwin/arm64 交叉编译（既有 flake 本轮未触发；同批 session_orchestrator.go 一度 501 行 → 抽出 session_tail.go）

internal/runtime/loop/request.go

@@ -8,22 +8,24 @@ import (
 	"git.yeij.top/AskaEth/Cyrene_v2/internal/runtime/model"
 )
 
-// buildRequest assembles a model.Request from the agent's public surface,
+// buildRequest assembles a model.Request from the turn's agent's public surface,
 // mirroring agent.buildRequest so the streaming path behaves identically to
-// RunStep (leading instruction message, tool schemas, model routing).
-func (l *Loop) buildRequest(ctx context.Context, messages []model.Message) model.Request {
+// RunStep (leading instruction message, tool schemas, model routing). The agent
+// and the reasoning knobs are resolved from ctx, so a session that switched roles
+// gets the new role's prompt and budget on its very next step (design/22 §4.2).
+func (l *Loop) buildRequest(ctx context.Context, messages []model.Message, ag *agent.Agent, knobs LoopKnobs) model.Request {
 	req := model.Request{
-		Model:           l.agent.Model(),
-		Messages:        l.assembleMessages(messages),
-		Temperature:     l.temperature(),
-		MaxTokens:       l.agent.MaxTokens(),
-		Purpose:         l.agent.Purpose(),
+		Model:           ag.Model(),
+		Messages:        assembleMessages(ag, messages),
+		Temperature:     temperatureOf(ag),
+		MaxTokens:       ag.MaxTokens(),
+		Purpose:         ag.Purpose(),
 		Tools:           l.toolSchemas(ctx, messages),
-		Thinking:        l.thinking,
-		ReasoningEffort: l.reasoningEffort,
+		Thinking:        knobs.Thinking,
+		ReasoningEffort: knobs.ReasoningEffort,
 	}
-	if l.agent.Router() != nil {
-		if m, _ := l.agent.Router().RouteRequest(req); m != "" {
+	if ag.Router() != nil {
+		if m, _ := ag.Router().RouteRequest(req); m != "" {
 			req.Model = m
 		}
 	}
@@ -32,8 +34,8 @@ func (l *Loop) buildRequest(ctx context.Context, messages []model.Message) model
 
 // assembleMessages prepends the agent's instructions as a leading system message
 // when they are non-empty, otherwise returns the input slice unchanged.
-func (l *Loop) assembleMessages(messages []model.Message) []model.Message {
-	instr := l.agent.Instructions()
+func assembleMessages(ag *agent.Agent, messages []model.Message) []model.Message {
+	instr := ag.Instructions()
 	if instr == "" {
 		return messages
 	}
@@ -43,10 +45,10 @@ func (l *Loop) assembleMessages(messages []model.Message) []model.Message {
 	return out
 }
 
-// temperature returns a pointer to the configured temperature, mapping zero to
-// nil so the provider default applies (matching the agent package).
-func (l *Loop) temperature() *float64 {
-	t := l.agent.Temperature()
+// temperatureOf returns a pointer to the agent's configured temperature, mapping
+// zero to nil so the provider default applies (matching the agent package).
+func temperatureOf(ag *agent.Agent) *float64 {
+	t := ag.Temperature()
 	if t == 0 {
 		return nil
 	}
@@ -140,14 +142,15 @@ const capacityRecallTopK = 5
 // habits and must never read owner-sensitive memory into a group or stranger intent
 // decision (design §2.2 红线: 不读陌生人敏感).
 func (l *Loop) capacityRecall(ctx context.Context, messages []model.Message) []capacity.RecallItem {
-	if l.retriever == nil || l.personaID == "" {
+	personaID := l.personaFor(ctx)
+	if l.retriever == nil || personaID == "" {
 		return nil
 	}
 	env := sessionEnvFrom(ctx)
 	if env.Restricted() {
 		return nil
 	}
-	results, err := l.retriever.Retrieve(ctx, l.personaID, queryFromMessages(messages), capacityRecallTopK)
+	results, err := l.retriever.Retrieve(ctx, personaID, queryFromMessages(messages), capacityRecallTopK)
 	if err != nil {
 		l.log.Warn("capacity recall", "error", err, "env", env.ActorEnv)
 		return nil