[Review] Reviewing #1570: Support for Support for SSE and NDJSON

[Olshansk]

Reviewing Streaming Support for SSE and NDJSON (used for LLMs) in #1750.

Next steps:

• Outlined in this comment: [RelayMiner] Streaming Support for SSE and NDJSON (used for LLMs)  #1570 (review)

Reproducing this locally (after running localnet):

make send_relay_path_REST
make pocketd_relayminer_relay_JSONRP

Changes made:

• Updated the LocalNet toolchain to make it easier to test REST & LLM services
• Added make pocketd_relayminer_relay_JSONRPC and revived make send_relay_path_REST
• Reviewing streaming RelayMiner changes in #1570 for code health & maintainbility
• Moved logic from sync.go into server.handleHttp to follow a similar pattern as server.handleHttpStream

---

The image below shows the result in LocalNet:

[RawthiL]

This PR seems to contain a total re-work of the code, should I review this new implementation instead of using the one I did?

[RawthiL]

I updated the base to latest main, so now this needs to be rebased so I can evaluate the logic changes (if any). I'm not sure if the streaming support was actually tested here. The images provided do not show relays performing streaming requests.

I think the approach of creating handlers per request (server.handleHttp/server.handleHttpStream) is correct. I have not implemented it so far because it makes rebases a hell..

I see lots of config changes but basically it seems that you are adding just two rules:

• send_relay_path_REST : Send a relay with path to Ollama (I cannot test if any of this works)
• pocketd_relayminer_relay_JSONRPC : Which sends a blockchain RPC to relayminer.
  I could add a third rule to make an RPC using pocketd relayminer relay to an overridden endpoint that can point to any OpenAI API compatibel endpoint. This way we will be able to test it without needing to deploy PATH or Ollama. Will that be enough?

P.S.: I see the header timeout removed, great..