Communication is key

Haha, beautiful solution:

Fix: new /deepsearch/stream (SSE) emitting a telemetry event every second while upstream runs, then result + done, mirroring the proven /deepresearch/stream. DeepSearchPanel now consumes
it; bytes flow from the first second, so no timeout can occur. /deepsearch/api stays for machine clients.

I could change molohttp server of course (it doesn't know that it's an option) but actually, i love this solution so much, i just keep it that way. It's actually top notch communication. If i would raise the time out on molohttp we would actually never find out if it really crashed or not. Now, we have guaranteees that the process is just doing it's thing.