Documentation fetch loop report & cache proposal

[ccsmart]

Greetings.
Just wanted to report on an observation.
I had a case where micode stepped into a loop of fetching (and evaluating) a sequence of ~5 documents from the web in what appeared to be an endless loop ( stopped it after about 10 rounds).
Request would therefore be for a self monitoring that looks for endless loops and intervenes.
Second is a proposal for local caching of documents fetched (ideally some per project logic so it can be efficiently trimmed) so if they are needed again, which i guess is not unlikely, they do not need to be fetched again.
Thanks for great micode and the offer to share it.

[vtemian]

Thanks for reporting this! The root cause was doom_loop: "allow" disabling OpenCode's built-in loop detection — it was set as collateral when all permissions were blanket-allowed, not intentionally.

Fix is up in #36: re-enables doom loop protection and adds a fetch-specific tracker that caches results, warns after 3 repeated fetches to the same URL, and hard-blocks after 5.