chore: single retry helper across services

[2ndtlmining]

Supersedes the code-quality half of #52 with the current state of the code.

There are now four separate retry implementations with different policies:

Location	Policy
cloudService.js retryApiCall()	2 attempts, 1s fixed delay
runningAppsProvider.js	3 attempts, 5s fixed delay
backupService.js withRetry()	attempts with backoff
hostLocationService.js	provider fallback, no retry

backupService.js's version is the best starting point. Extract it to src/lib/retry.js and use it everywhere, keeping per-caller attempt counts and delays as arguments rather than flattening every call site to one policy.

While doing it, cloudService.js and wordpressService.js still use console.log/console.error in places and bypass the pino logger introduced in #36.

Related

• Retry logic and failure handling are inconsistent across near-identical services #52 (the original report)
• External API fetches aren't protected by the circuit breaker #55 (external API calls have no circuit breaker — a shared helper is the natural place to add one)