feat(thermal): Add a standalone thermal/safety run, and move margin out of the soak loop

4_Blanton_Script_thermal_safety.ttl brings the DUT up the way Script A
does, then loads it - mlucas-avx2 -cpu 0:15 in the background plus
blanton_tr518.sh start - and samples platform data and both switch
boards' rail power every minute. Power per rail under sustained full
load, over hours, is the whole point of it.

It is standalone: nothing starts it, it starts nothing, and it must not
share a DUT with Script B because both take over the cd ports.

It also has no teardown. When you interrupt it the unit is still in
port cd lb=mac / l2 learn off / test mode nr=yes and mlucas is still
running, and show interfaces status does not reveal any of that. Called
out in the version history, the release notes, CLAUDE.md gotchas and the
ARCHITECTURE future-work list, because whoever uses the box next will not
find out on their own.

Script B's per-round loop drops the margin scan. The loop is meant to be
a quick minute-by-minute sample and margin_status_all.sh walks nine
LTC2980s over I2C, which dominated the round. Script C takes it instead,
once, before show uptime. The trade-off is real: margin drift DURING the
soak is no longer visible, only a start and an end reading. Documented
rather than glossed over, with a suggestion (scan every N rounds) if it
needs to come back.

Script C also sends exit and waits for "Script done" before
hw-test-session finish, so finish runs in the login shell rather than
inside the session it is closing.

Adds docs/Blanton_Test_Flow.drawio: all four scripts side by side, with
the config flags that change each step marked, Script B's one real goto
drawn, and the A->B->C handoffs shown as the manual steps they are.

The new script arrived LF-only while every other .ttl is CRLF and
.gitattributes marks *.ttl as -text, so what is committed is exactly what
Tera Term reads. Converted.

Script A -> V1.1.4.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014RetWKFZFG1ZHcitQAyhwM

CLAUDE.md

@@ -9,7 +9,7 @@ DUT 端的取數能力由一組 **bash 工具庫** 提供（CB FPGA `pcimem` →
 
 完整架構見 [ARCHITECTURE.md](ARCHITECTURE.md)。
 
-**目前狀態（2026-08-30）**：Script A 為 **V1.1.3**（A 集中記錄 A/B/C 三支的變更）。涵蓋
+**目前狀態（2026-08-31）**：Script A 為 **V1.1.4**（A 集中記錄所有 TTL 的變更）。涵蓋
 PCIe AER / DDR EDAC / PMON+TPM / 電壓 margin（9 顆 LTC2980）/ SWB loopback 線速流量 /
 10G+1G 管理網路 ping / BMC DDR 與 USB / NVMe 健康，壓力層走平台的 `bgctl`。
 switch unit 數量由 `config.ttl` 的 `SWB_UNIT0` / `SWB_UNIT1` 宣告，風扇轉速由 `FAN_SPEED` 指定
@@ -23,7 +23,10 @@ TR518 負載（`blanton_tr518.sh`，**TTL 尚未接**）、SWB 軌功耗（`TH6_
 Script A 開頭會先收掉前一輪殘留（`hw-test-session finish` + `bgctl reset/stop --all`），
 再 `bgctl list` 確認沒有 job 留著佔資源。
 NFC tag 讀取由 `nfc_polling.sh` 背景輪詢（A 做基線、B 起、C 停並 report）。
-最後一次發布是 **V1.1.2**；SWB 的 I2C 通道與位址已於 2026-08-19 上機驗證。
+另有 **`4_Blanton_Script_thermal_safety.ttl`**（熱 / 安規，**獨立跑不接 A→B→C**，滿載後每分鐘
+取 PMON + 兩片 SWB 的軌功耗；**沒有收尾段**，中斷後要手動 `blanton_tr518.sh stop`）。
+流程圖：`docs/Blanton_Test_Flow.drawio`。
+最後一次發布是 **V1.1.3**；SWB 的 I2C 通道與位址已於 2026-08-19 上機驗證。
 margin 的 `VNOM` 已於 2026-08-30 改為 Blanton 實測中位數（144 條軌中 86 條上調 0.8~2.5%）。
 交付用 `./publish/publish.sh` 打包。
 
@@ -219,6 +222,11 @@ Tera Term 連 COM port (115200-8-N-1)
 - ⚠️ **ICB/PDB 一律走 VI2C（`blanton_icb_vi2c.sh`），不要走 VSPI**：VSPI ~18 ms/reg 會讓 I2C
   byte 相間隔拉到 ~50 ms，踩中 SMBus 25–35 ms timeout，裝置直接 NACK。實測全掃 9.4 s vs 34.0 s。
 - ⚠️ **I2C 讀一律 Repeated START**（不是 STOP + START），否則 PMBus/SMBus 裝置（VRM、EFUSE）不理。
+- ⚠️ **Script 4（thermal/safety）沒有收尾**：`while 1` 之後就是檔尾。中斷後 DUT 停在
+  `port cd lb=mac` / `l2 learn off` / `test mode nr=yes`，`mlucas-avx2` 也還在背景。
+  下一個測試會在被污染的狀態上開始，而且 **`show interfaces status` 看不出來**。
+  收尾：`blanton_tr518.sh stop` + `kill $(jobs -p)`
+- ⚠️ **Script 4 與 Script B 不能共用同一台 DUT**：兩者都會接管 `port cd` 的 loopback
 - ⚠️ **`settings/*.conf` 的 `VNOM` 是實測中位數，不是規格值**（2026-08-30 起）。它同時是
   `margin_status` 偏差 % 的分母、也是 `margin_set` 推算 high/low/OV/UV 的基準，所以改它會
   同時移動「看到的數字」和「實際打出去的電壓」。**跨這個日期的 margin log 不能直接互比。**