Core: Add manifest writer pool to avoid commit delays

[xxubai]

The default iceberg-worker-pool thread pool is shared by many tasks or services. During snapshot writes, if the thread pool is saturated, it may result in significant delays in committing operations.
On the other hand, retries caused by manifest write failures may occupy the iceberg-worker-pool, potentially blocking other tasks that rely on the same thread pool.

[singhpk234]

Thank you so much for you contribution !

Have some doubts / questions :

• can you please elaborate more about your workload (a before vs after would be really appreaciated) ?
• what the worker pool threads are busy doing ?
• is it in a situation even if you increase the pool size, you will not be able to get it for your actions ?