feat: provider-aware rate limit reset delay

Previously getRateLimitResetDelayMs only read the Anthropic-specific
'anthropic-ratelimit-unified-reset' header (Unix timestamp), returning
null for every other provider. This meant OpenAI, GitHub, and Codex
users in persistent retry mode (CLAUDE_CODE_UNATTENDED_RETRY=1) always
fell back to dumb exponential backoff even when the server included an
exact reset time in the response headers.

This change makes the function provider-aware:

- firstParty (Anthropic): existing behaviour preserved — reads
  'anthropic-ratelimit-unified-reset' Unix timestamp
- openai / codex / github: reads 'x-ratelimit-reset-requests' and
  'x-ratelimit-reset-tokens' (OpenAI relative duration strings like
  "1s", "6m0s", "1h30m0s"), picks the larger of the two so retries
  don't fire before both token and request limits have reset
- bedrock / vertex / foundry / gemini: returns null (no standard
  reset header for these providers)

Adds parseOpenAIDuration() as an exported helper to convert OpenAI's
duration format into milliseconds.

16 new tests covering all provider paths and edge cases.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

src/services/api/withRetry.ts

@@ -11,7 +11,7 @@ import { isAwsCredentialsProviderError } from 'src/utils/aws.js'
 import { logForDebugging } from 'src/utils/debug.js'
 import { logError } from 'src/utils/log.js'
 import { createSystemAPIErrorMessage } from 'src/utils/messages.js'
-import { getAPIProviderForStatsig } from 'src/utils/model/providers.js'
+import { getAPIProvider, getAPIProviderForStatsig } from 'src/utils/model/providers.js'
 import {
   clearApiKeyHelperCache,
   clearAwsCredentialsCache,
@@ -811,12 +811,49 @@ function getRetryAfterMs(error: APIError): number | null {
   return null
 }
 
-function getRateLimitResetDelayMs(error: APIError): number | null {
-  const resetHeader = error.headers?.get?.('anthropic-ratelimit-unified-reset')
-  if (!resetHeader) return null
-  const resetUnixSec = Number(resetHeader)
-  if (!Number.isFinite(resetUnixSec)) return null
-  const delayMs = resetUnixSec * 1000 - Date.now()
-  if (delayMs <= 0) return null
-  return Math.min(delayMs, PERSISTENT_RESET_CAP_MS)
+/**
+ * Parse OpenAI-style relative duration strings into milliseconds.
+ * Formats: "1s", "6m0s", "1h30m0s", "500ms", "2m"
+ * Returns null for unrecognized formats.
+ */
+export function parseOpenAIDuration(s: string): number | null {
+  if (!s) return null
+  // Try matching hours/minutes/seconds/milliseconds components
+  const re = /^(?:(\d+)h)?(?:(\d+)m(?!s))?(?:(\d+)s)?(?:(\d+)ms)?$/
+  const m = re.exec(s)
+  if (!m || m[0] === '') return null
+  const h = parseInt(m[1] ?? '0', 10)
+  const min = parseInt(m[2] ?? '0', 10)
+  const sec = parseInt(m[3] ?? '0', 10)
+  const ms = parseInt(m[4] ?? '0', 10)
+  const total = h * 3_600_000 + min * 60_000 + sec * 1_000 + ms
+  return total > 0 ? total : null
+}
+
+export function getRateLimitResetDelayMs(error: APIError): number | null {
+  const provider = getAPIProvider()
+
+  if (provider === 'firstParty') {
+    const resetHeader = error.headers?.get?.('anthropic-ratelimit-unified-reset')
+    if (!resetHeader) return null
+    const resetUnixSec = Number(resetHeader)
+    if (!Number.isFinite(resetUnixSec)) return null
+    const delayMs = resetUnixSec * 1000 - Date.now()
+    if (delayMs <= 0) return null
+    return Math.min(delayMs, PERSISTENT_RESET_CAP_MS)
+  }
+
+  if (provider === 'openai' || provider === 'codex' || provider === 'github') {
+    const reqHeader = error.headers?.get?.('x-ratelimit-reset-requests')
+    const tokHeader = error.headers?.get?.('x-ratelimit-reset-tokens')
+    const reqMs = reqHeader ? parseOpenAIDuration(reqHeader) : null
+    const tokMs = tokHeader ? parseOpenAIDuration(tokHeader) : null
+    if (reqMs === null && tokMs === null) return null
+    // Use the larger delay so we don't retry before both limits reset
+    const delayMs = Math.max(reqMs ?? 0, tokMs ?? 0)
+    return Math.min(delayMs, PERSISTENT_RESET_CAP_MS)
+  }
+
+  // bedrock, vertex, foundry, gemini — no standard reset header
+  return null
 }