Fix request headers not fowarded in fetch_upstream

[Jdyn]

Introduces put_request_headers to merge conn headers into the request before making upstream fetch calls. Also renames put_headers to put_conn_headers for clarity.

Closes #105

I found a few ways to approach this

One way I considered was to just override the response headers after we receive the response with the original headers with the CORS plug:

defp put_headers(conn, headers) do
  headers
  |> Map.delete("transfer-encoding")
  |> Enum.reduce(conn, fn {header, values}, conn ->
    Enum.reduce(values, conn, fn value, conn ->
      Plug.Conn.put_resp_header(conn, header, value)
    end)
  end)
+  |> Phoenix.Sync.Plug.CORS.call()
end

Depends on what behavior you guys consider correct, But I assumed its more correct to merge the client request headers with the original conn request headers before sending the request, since Electric handles the headers logic upstream anyway.

I spent a lot of time trying to get a test setup for this, but couldn't figure out how to test sync_render in http mode. My tests weren't going to the paths I expected when using the sync_render in the Sandbox. Feel free to modify stuff.

[Copilot]

Pull Request Overview

This PR fixes an issue where request headers from the client connection were not being forwarded to upstream Electric requests in the fetch_upstream function. The change ensures client headers are properly merged with the request before making upstream calls.

• Adds a new put_req_headers function to merge client connection headers into the request
• Renames put_headers to put_resp_headers for better clarity between request and response header handling
• Updates the fetch_upstream flow to include client headers in upstream requests

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[magnetised]

@Jdyn I was about to have a proper look at this -- haven't had much time this week, other priorities interfering. Why did you close?

[Jdyn]

Oh, I had just figured you would have your own sol. Will re-open but also not entirely sure how elegant this solution is. Happy to make any changes to get it in. I have been using the branch in prod and it fully works for me.

[magnetised]

Oh, I had just figured you would have your own sol.

I wondered what your reasons for closing were - whether the patch didn't work or something.

@Jdyn rather than request a bunch of changes plus a test I did end up making my own pr based on this: #107 -- your basic approach seemed valid, I just tweaked to handle compatibility with the elixir client's handling of headers (which on review isn't great tbh).

[magnetised]

Thanks for the PR and the issue. Closing in favour of #107