Please document how upstream and fallback work

1. How are upstream servers used? Random, round-robin, fastest ... (`--all-servers` is clear)

2. What does upstream not available mean? dnsproxy cannot connect to the server, or it can but gets back a DNS error?

3. When the upstreams are not available, how are the fallback servers used? In order, random, ...

4. Does dnsproxy monitor the upstream servers and switches back to them from fallkback, when available again?

No files, tests, or entry points are named. Start by locating the code that selects upstream and fallback DNS servers, then trace how unavailable servers and recovery are handled. Done means documenting selection order, failure semantics, fallback behavior, and switching back to upstreams when they recover.

Escrita pelo modelo de indexação a partir do texto da issue.