chore(caveman): purge plugin + always-on integration

Disable + uninstall caveman@caveman and delete every repo dependency on
it: SessionStart/UserPromptSubmit hook blocks, standalone hook files,
settings.json enabledPlugins + marketplace entries, install-plugins.sh
STEP 5.5, update-all.sh refresh step, plugins.lock.json entry, doctor.sh
checks, lib/detect-plugins.sh helpers, lib/profile.sh + plugin-advisor +
skills/profile protected-list entries, .gitignore runtime-file block,
and README/USAGE docs. Dead /caveman:compress refs replaced with
manual/claude.ai guidance. Memory-registry terse-format convention kept
(separate subsystem). Version 3.4.0 -> 3.5.0.

On a subscription plan caveman's ~75% output-token compression has no
cost benefit, and the always-on hooks added friction on validation
gates and client deliverables.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X3e8LaH2vymmxyh36h3jFU

agents/plugin-advisor.md

@@ -376,7 +376,7 @@ gstack + managed plugins — sessions stay focused and passive token cost drops.
 
 `profile set <name>` actually toggles plugins (`claude plugin enable|disable`)
 and MCPs (delegates to `lib/toggle-external.sh` for `magic`) — not just
-advisory. Always-on plugins (`caveman`, `security-guidance`, `superpowers`)
+advisory. Always-on plugins (`security-guidance`, `superpowers`)
 are protected. Managed plugins that `set` may toggle:
 `ui-ux-pro-max@ui-ux-pro-max-skill`, `plugin-dev@claude-code-plugins`,
 `pr-review-toolkit@claude-code-plugins`. Other plugins are never auto-toggled.