feat: add retry with exponential backoff for transient API failures

[iamfj]

Problem

GraphQLClient and FileService have zero retry logic. Linear's API has rate limits and can return transient 5xx errors. Currently:

• HTTP 429 (rate limit) → cryptic GraphQL error, no indication it's a rate limit
• HTTP 5xx (server error) → one-shot failure, no retry
• Network hiccups → immediate failure

This is especially painful for issues create which may have resolved 4-5 entity IDs via separate calls before the final mutation fails — all that work is lost.

Proposed solution

Add a shared retry utility and apply it to GraphQLClient.request():

// src/common/retry.ts
export async function withRetry<T>(
  fn: () => Promise<T>,
  options?: { maxRetries?: number; baseDelayMs?: number }
): Promise<T> {
  const { maxRetries = 3, baseDelayMs = 500 } = options ?? {};
  for (let attempt = 0; attempt <= maxRetries; attempt++) {
    try {
      return await fn();
    } catch (error) {
      if (attempt === maxRetries || !isRetryable(error)) throw error;
      const delay = baseDelayMs * 2 ** attempt;
      await new Promise((r) => setTimeout(r, delay));
    }
  }
  throw new Error("unreachable");
}

Retry behavior

Status	Retry?	Reason
429	✅	Rate limited — wait and retry
500-599	✅	Transient server error
401/403	❌	Auth error — retrying won't help
400, 404	❌	Client error — request is wrong
Network error	✅	Transient connectivity

Where to apply

• GraphQLClient.request() — wrap the rawRequest call
• FileService fetch calls — wrap each fetch()

Acceptance criteria

• withRetry() utility with exponential backoff exists
• GraphQLClient.request() retries on 429/5xx (max 3 attempts)
• Auth errors are NOT retried
• Retry delay follows exponential backoff (500ms, 1s, 2s)
• Unit tests for retry logic (success after retry, exhausted retries, non-retryable errors)
• Existing tests still pass