fix(compaction): configurable summarizer tool_choice; wider overflow detection

Port of upstream grok-build 0.2.105 + 0.2.118.

- CompactionToolChoice (auto default | none), resolved env
  KIGI_COMPACTION_TOOL_CHOICE > [features].compaction_tool_choice >
  remote settings. Some OpenAI-compatible servers reject
  tool_choice:"none" with tools attached, failing every long-session
  compaction; auto keeps them working (the summarization prompt already
  forbids tool use). Threaded from agent + subagent spawn through
  CompactionConfig into generate_session_compact; the Responses gate
  counts hosted tools too, so a hosted-only request still carries the
  configured choice.
- is_context_length_error recognizes the "current message ... exceeds
  budget" phrasing, so that overflow is Fatal instead of burning the
  retry budget on a deterministic failure.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01KSy9S31pZNWXtRZNVJ1sDf

crates/codegen/kigi-config-types/src/lib.rs

@@ -668,6 +668,11 @@ pub struct RemoteSettings {
     /// remote settings verbatim-input flag; env (`KIGI_COMPACTION_VERBATIM_INPUT`) and config override it. `None` = default (true).
     #[serde(default)]
     pub compaction_verbatim_input: Option<bool>,
+    /// Compaction summarizer `tool_choice` (`auto` | `none`) from remote
+    /// settings. Env (`KIGI_COMPACTION_TOOL_CHOICE`) and config override it.
+    /// Some backends reject `tool_choice: "none"` with tools attached.
+    #[serde(default)]
+    pub compaction_tool_choice: Option<String>,
     /// Master switch for jemalloc heap sampling + threshold dumps.
     /// `Some(true)` enables, `Some(false)` kill-switch, `None` = client default off.
     #[serde(default)]