docs: add warning and troubleshooting for self-hosted LLM provider config bug (#1577)

[Adityakk9031]

Resolves #1577.

Description

Adds a troubleshooting section to the self-hosted configuration documentation, explaining the root cause and workaround for the [llm] No model provider configured\ error on document processing.

Root Cause

During binary compilation of \supermemory-server\ with Bun (\�un build --compile), environment variables like \process.env.OPENAI_API_KEY\ are evaluated and inlined at compile/build time (resolved to \undefined\ or default values on CI/CD runner environments).
At runtime:

1. The memoized setup module (\s55()) runs exactly once when server modules are loaded and uses these inlined references, causing the internal LLM registry to construct model factories with
   ull\ keys.
2. Inside sandboxed background execution environments (like Miniflare/workerd running the \IngestContentWorkflow), \process.env\ is not propagated from the host process unless explicitly configured as bindings.
3. This creates a disconnect where \supermemory-server doctor\ reports successful configuration (as it connects to the database store directly), but runtime memory generation crashes.

Workaround Documented

• Always launch the server using the generated wrapper script (~/.local/bin/supermemory-server), which properly exports keys from ~/.supermemory/env\ using \set -a.
• Explicitly inject keys into the parent environment if using process managers (systemd/PM2) or Docker.