fix(claude): add interleaved-thinking beta header, AMP gzip error decoding, normalizeClaudeBudget max_tokens

1. Always include interleaved-thinking-2025-05-14 beta header so that
   thinking blocks are returned correctly for all Claude models.

2. Remove status-code guard in AMP reverse proxy ModifyResponse so that
   error responses (4xx/5xx) with hidden gzip encoding are decoded
   properly — prevents garbled error messages reaching the client.

3. In normalizeClaudeBudget, when the adjusted budget falls below the
   model minimum, set max_tokens = budgetTokens+1 instead of leaving
   the request unchanged (which causes a 400 from the API).

internal/thinking/provider/claude/apply.go

@@ -194,7 +194,9 @@ func (a *Applier) normalizeClaudeBudget(body []byte, budgetTokens int, modelInfo
 	}
 	if minBudget > 0 && adjustedBudget > 0 && adjustedBudget < minBudget {
 		// If enforcing the max_tokens constraint would push the budget below the model minimum,
-		// leave the request unchanged.
+		// increase max_tokens to accommodate the original budget instead of leaving the
+		// request unchanged (which would cause a 400 error from the API).
+		body, _ = sjson.SetBytes(body, "max_tokens", budgetTokens+1)
 		return body
 	}