fix(billing): apply margin multiplier to subscription quota consumption

Subscription plans now deduct period_tokens_used as upstream tokens ×
marginMultiplier, aligning package quota burn with wallet cost-mode pricing.

Co-authored-by: Cursor <cursoragent@cursor.com>

user-auth.mjs

@@ -3,7 +3,12 @@ import fs from 'node:fs';
 import net from 'node:net';
 import path from 'node:path';
 import { Algorithm as Argon2Algorithm, hashRawSync as argon2HashRawSync } from '@node-rs/argon2';
-import { computeDeltaCostCents, loadBillingConfig, normalizeTokenState } from './billing.mjs';
+import {
+  computeDeltaCostCents,
+  computeSubscriptionQuotaTokens,
+  loadBillingConfig,
+  normalizeTokenState,
+} from './billing.mjs';
 import { buildInsufficientBalancePayload, loadRechargeConfig } from './billing-recharge.mjs';
 import {
   buildAgentExtensionPolicy,
@@ -1444,6 +1449,7 @@ export function createUserAuth(pool, options = {}) {
         tokenState.accumulatedOutputTokens - Number(previous?.lastOutputTokens ?? 0),
       );
       const deltaTokens = deltaIn + deltaOut;
+      const subscriptionQuotaTokens = computeSubscriptionQuotaTokens(deltaIn, deltaOut, config);
 
       await conn.query(
         `INSERT INTO h5_session_billing_state
@@ -1467,7 +1473,7 @@ export function createUserAuth(pool, options = {}) {
       // Subscription quota check: consume tokens from active plan before touching balance.
       let subscriptionCovered = false;
       if (costCents > 0 && subscriptionService) {
-        const coverage = await subscriptionService.consumeQuota(userId, deltaTokens, conn);
+        const coverage = await subscriptionService.consumeQuota(userId, subscriptionQuotaTokens, conn);
         if (coverage.fullyCovers) {
           costCents = 0;
           subscriptionCovered = true;