fix(model-error-classifier): scope forbidden-provider retry to specific phrase

Bare "403" and "forbidden" substring patterns (added in 034744cb to
retry the "Selected provider is forbidden" case from PR #3706) matched
any error message containing those tokens — tool-level 403s, file-perm
"forbidden" messages, unrelated upstream errors that happened to spell
"forbidden". The legacy model-fallback path then armed setPendingModelFallback
on those unrelated errors, jumping Sisyphus to claude-opus-4-7 (first
entry of its fallback chain) regardless of the user's configured model.

Replace the bare patterns with the specific phrases PR #3706 actually
targeted, and add regression tests asserting unrelated 403/forbidden
messages stay non-retryable.

Reported-by: ilove_borshch on Discord (#omo-help)

src/shared/model-error-classifier.test.ts

@@ -248,6 +248,28 @@ describe("model-error-classifier", () => {
     //#then
     expect(result).toBe(true)
   })
+
+  test("does not treat unrelated forbidden messages as retryable", () => {
+    //#given
+    const error = { message: "EACCES: forbidden write to /etc/hosts" }
+
+    //#when
+    const result = shouldRetryError(error)
+
+    //#then
+    expect(result).toBe(false)
+  })
+
+  test("does not treat unrelated 403 messages as retryable", () => {
+    //#given
+    const error = { message: "Tool returned HTTP 403 for the requested URL" }
+
+    //#when
+    const result = shouldRetryError(error)
+
+    //#then
+    expect(result).toBe(false)
+  })
 })
 
 export {}

src/shared/model-error-classifier.ts

@@ -72,8 +72,8 @@ const RETRYABLE_MESSAGE_PATTERNS = [
   "504",
   "429",
   "529",
-  "403",
+  "selected provider is forbidden",
-  "forbidden",
+  "provider is forbidden",
 ]
 
 /**