Self-hosting: document running supermemory-server with Docker

[Ansh-Sonkusare]

The self-hosting docs mention Docker (the "Docker / non-interactive" tip in the quickstart), but there's no Dockerfile, compose file or image anywhere, so everyone running supermemory-server in a container writes their own.

I've been running v0.0.8 in Docker and Podman for a while (supermemory-docker, plus a NixOS module). A few things caught me out, and a docs page would save people from them:

• There's no wizard without a TTY, so an LLM key has to come from env or the server won't start.
• /data needs to be a volume, or the API key and database are gone after a rebuild.
• supermemory-server upgrade doesn't make sense inside a container. You rebuild with a new version instead.
• Inside the container, localhost isn't the host, so pointing at Ollama or LM Studio needs host.docker.internal (plus host-gateway on Linux).
• The localhost auto-auth still works through a published port on the host. Requests from other machines or other containers get a 401 unless they send the key.

Proposal

1. Docs: a self-hosting/docker page with a Dockerfile (pinned version, sha256-verified against the release .sha256, non-root) and a compose file. I'll open a PR for this. I've built and run everything on it against v0.0.8.
2. Maybe later: an official image published to GHCR alongside each server-v* release. The release pipeline isn't in this repo, so that's your call. I'm happy to help if you want it.

[linear-code]

ENG-1283

[Ansh-Sonkusare]

Following up on trust in the binary: I opened #1737 asking for signed release builds with build provenance (GitHub artifact attestations or Sigstore). That would let the Dockerfile on the docs page verify who built the binary, not just its checksum. It doesn't depend on #1299.