Make the internal polling interval configurable

[dolfim-ibm]

I may be wrong and it's unrelated but there's a 5s hardcoded polling interval in

docling-serve/docling_serve/app.py

Line 297 in 0d4545a

await asyncio.sleep(5)

which explains the 5s overhead when using synchronous endpoints.
If I'm correct. @dolfim-ibm would you mind reopening the issue to make the pollig interval configurable?

Originally posted by @prein in #205