chore(release): use effort: high (unambiguously supported) instead of xhigh

xhigh's support varies by model (gpt-5.5 may coerce it to high anyway), so use
high directly to remove the ambiguity. The diff is injected into the prompt
regardless, so effort only affects reasoning quality, not whether the model
sees the code.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Author: yanxue06

.github/blocks/bump-monorepo-versions/action.yaml

@@ -135,9 +135,9 @@ runs:
         openai-api-key: ${{ inputs.openai-api-key }}
         safety-strategy: read-only
         # Diff-based breaking-change judgment is a real reasoning task — run at
-        # max effort. The diff is provided per package below; the agent may also
+        # high effort. The diff is provided per package below; the agent may also
         # run read-only git for any truncated file.
-        effort: xhigh
+        effort: high
         prompt: |
           You are a release manager for the "${{ inputs.service-name }}" monorepo.
           Prerelease: ${{ inputs.prerelease }}

.github/blocks/determine-publish-version/action.yaml

@@ -131,9 +131,9 @@ runs:
         openai-api-key: ${{ inputs.openai-api-key }}
         safety-strategy: read-only
         # Judging backward-compatibility from a diff is a real reasoning task —
-        # run it at max effort. This step runs once per release and a wrong major
-        # bump is expensive, so the budget is justified.
-        effort: xhigh
+        # run it at high effort. This step runs once per release and a wrong
+        # major bump is expensive, so the budget is justified.
+        effort: high
         prompt: |
           Determine the next semantic version after ${{ steps.last-release.outputs.version }}.
           Current version: ${{ steps.last-release.outputs.version }}