ci(mirror): 单资产 cap 3min + 镜像段 10min 硬顶

按 review 定档:
- `MIRROR_UPLOAD_TIMEOUT` 120 → **180s**。mcpp 最大资产约 30MB(单包不超 100M),
  3 分钟还没传完不是「慢」是「卡住」,不值得继续占 CI —— 跳过并由本地补传
  (gate 会打印确切的补传命令)。
- Mirror 步骤加 **`timeout-minutes: 10`** 硬顶。健康的 run 远在此之内;10min 是
  兜底,防止某一端病态时烧光 job 预算(v0.0.94 就是 30min 挂死且零可见性)。

Author: Sunrisepeak

.github/tools/mirror_res.sh

@@ -55,9 +55,11 @@ info() { echo "[mirror] $*"; }
 # completeness gate at the bottom is still the pass/fail, so a skipped asset
 # fails the release loudly instead of silently shipping a half mirror.
 #
-# Raise it (or set it per-host) when the cross-border leg legitimately needs
-# longer than the cap for the big archives; the gate will tell you.
-: "${MIRROR_UPLOAD_TIMEOUT:=120}"
+# Sizing: mcpp's largest asset is ~30MB (no package exceeds 100MB). 180s is a
+# generous ceiling for that — an upload still running at 3min is not "slow", it
+# is stuck, and the right move is to stop paying CI for it and push that one
+# asset by hand (the gate below prints the exact command).
+: "${MIRROR_UPLOAD_TIMEOUT:=180}"
 
 DL="$(mktemp -d)"; trap 'rm -rf "$DL"' EXIT
 

.github/workflows/release.yml

@@ -725,6 +725,13 @@ jobs:
 
       - name: Mirror binaries to xlings-res/mcpp (gh + gtc)
         if: ${{ env.XLINGS_RES_TOKEN != '' }}
+        # Hard ceiling for the whole mirror segment. The script caps each asset
+        # at MIRROR_UPLOAD_TIMEOUT (180s) and abandons anything slower, so a
+        # healthy run lands well inside this; 10min is the backstop that keeps a
+        # pathological host from burning the job's budget the way v0.0.94 did
+        # (30min, killed, zero per-asset visibility). Whatever gets skipped is
+        # reported by name + size and pushed by hand.
+        timeout-minutes: 10
         env:
           GH_TOKEN: ${{ secrets.XLINGS_RES_TOKEN }}
         run: |