Scheduling: @Scheduled, ShedLock, Retry

Process in bounded chunks with per-chunk transactions, flush+clear the persistence context for memory, record a checkpoint for resumability, throttle the database, and skip-and-log bad records instead of aborting.

`@Retryable` re-invokes a method on transient failures with configurable backoff, `@Recover` is the fallback when retries are exhausted, and `@EnableRetry` activates the feature through AOP proxies.

ShedLock acquires a named lock in a shared store so only one instance runs the job; leader election makes one node the leader; Quartz clustering assigns triggers to exactly one node via a shared database.

`@Scheduled` is single-threaded by default, has no persistence of triggers, and has no runtime rescheduling -- fix with a thread pool, Quartz, and programmatic dynamic triggers respectively.