fix: put the tier back where "on" by itself is not on

The minimax wire lost its tier value on the assumption that these models
reason by default, so injecting nothing still reads as "on". T10 measured
the real gateway and the assumption does not hold: MiniMax-M3 with no
reasoning parameter did not reason in 5 of 5 rounds, while all six
strength values worked. Existing downstreams on ENABLE_THINKING=true
went from reasoning to silently not reasoning, and the capability table
cannot catch it because phase 5 lets auto through unconditionally.

Restore on_base to the old {"reasoning_effort": "medium"} verbatim. This
is a stopgap - it hands the tier choice back to the library, which this
work set out to remove. The real fix is to constrain auto by the
capability table, a public behaviour change tracked as issue #21.

The assertions that said "minimax injects no tier on the on-tier" go
back with it; each carries a note on why it moved twice.

tests/e2e/test_thinking_live.py

@@ -771,9 +771,10 @@ async def _anchor_off_against_on(
 
     锚点取 `completion_tokens` 的相对比较(关闭档最大值 < 开启档最小值),**不含
     任何魔数**: 推理段计在 completion 里,真开着时两档差一个数量级(实测 kimi-k3
-    关闭档恒 9 token)。取 `max` 档而非 `auto`: 后者对 minimax 一路等于"什么都不注入"
-    (`on_base={}`),那是模型默认档而不是"开",拿它当对照组会把 M3 这种默认不推理的
-    模型判成"分不开"。`max` 打不通时才退到 `auto`。
+    关闭档恒 9 token)。取 `max` 档而非 `auto`: 后者在 `on_base={}` 的 provider 上
+    等于"什么都不注入",那是模型默认档而不是"开",拿它当对照组会把 M3 这种默认不推理的
+    模型判成"分不开"(minimax 段已按 issue #21 改回带 medium,openai/anthropic/google
+    三段仍是空片段,故该风险仍在)。`max` 打不通时才退到 `auto`。
     """
     off_usable = [o for o in off_observations if _probe_ok(o)]
     for tier in (Effort.MAX, Effort.AUTO):