fix(consult): record a tool call the tool never received

staging 手测 run 951a841e 第一次工具调用发出 tool.failed 后重试成功，但回执里
steps 只有 skill 与那次成功的 tool，stepBudget.used 为 2——失败的那次完全不存在。
客户端看见失败过一次，回执说没有，两边都查不到为什么。

工具的 inputSchema 是 strict 的，模型参数不合法时 Mastra 在调用 execute 之前就拒了，
于是工具体内一切都没跑：调用不计数、失败步不记录、连 chart-calculation 活动事件都没
发出（这也是本次定位的证据——失败那次没有任何 activity，重试那次有）。工具无法记录
一次它从未收到的调用。叠加两处：safeToolError 把非超时非取消的错误全塌成
calculation_failed，而即使失败落进工具体的 catch，consultationWorkflowFailureCode
对非 ConsultationWorkflowError 返回 undefined、append 处又写成可选省略，于是最需要
解释的那条记录恰好是唯一没有原因的记录。

改为在流层补记：流是唯一能观测到全部工具失败的位置，无论失败在 schema 这侧还是
execute 那侧，且它持有 startedAt 因而能给出时长。tool-error 分支比对「流已见的错误数」
与「state 里已有的失败 tool 步数」，只在前者更多时补一条，工具仍记录它能看见的失败，
两者不重复计。另新增 consultationToolFailureCode，令每个错误都解析出一个码。

failureCode 刻意仍不进公开回执：白名单与「the public receipt never carries the
internal failure classification」是刻意约束，workflow_rate_limited 这类后端内情不该
上线到客户端。原因走可观测日志的 toolCalls[].failureCode。客户端能看到「有一步失败」，
运维能在日志里看到为什么。

另记入 BUG-268 的线上实测值：单领域 referenceReads 两次均为 2，多领域为 0——不是
从不读方法，而是最需要方法的多领域路径一份都没打开。

Co-authored-by: Cursor <cursoragent@cursor.com>

scripts/jyotish_api_server.py

@@ -699,7 +699,7 @@ def _attach_local_consultation_layers(handler, chart: dict, birth_payload: dict,
 #
 # The gate must also not be weaker than what the product tells the user it will use. The frontend
 # domain registry declares each domain's layers for its evidence preview, and every entry there that
-# names a real section is required here too; a test pins that intersection (BUG-269). The registry
+# names a real section is required here too; a test pins that intersection (BUG-270). The registry
 # additionally names human-facing items ('7th house/lord', 'negative holdout gate') and D11, which
 # the engine does not build at all — those are deliberately not requirements.
 _ROUTE_REQUIRED_LAYERS = {

scripts/run_quality_gate.py

@@ -53,7 +53,7 @@ CORE_PYTEST_TARGETS = [
     "tests/test_external_oracle_sanity_closure.py",
     # The staging gate runs the quick profile, so a guard absent from this list never runs in CI.
     # These pin the answer-truth contract every product consultation is built on, and their failure
-    # mode is silent widening — nothing errors when they regress (BUG-267, BUG-269).
+    # mode is silent widening — nothing errors when they regress (BUG-267, BUG-270).
     "tests/test_consultation_consumer_context.py",
 ]