fix: drop stale "Global Memory" instruction from main seed prompt

groups/main/CLAUDE.md tells the main agent it "can read and write to
/workspace/global/CLAUDE.md for facts that should apply to all groups."
That mount no longer exists: the c8fc1da cutover moved the universal base
into container/CLAUDE.md (RO at /app/CLAUDE.md) and groups/global/ is deleted
on every startup, so the path never mounts.

There is no runtime-writable global-memory path to repoint to — per-group
memory lives in CLAUDE.local.md, and the shared base is build-time and
read-only — so the section is removed rather than rewritten.

Related: #2312 (same incomplete c8fc1da cleanup).

Author: CutSnake01

groups/main/CLAUDE.md

@@ -257,12 +257,6 @@ Read `/workspace/project/data/registered_groups.json` and format it nicely.
 
 ---
 
-## Global Memory
-
-You can read and write to `/workspace/global/CLAUDE.md` for facts that should apply to all groups. Only update global memory when explicitly asked to "remember this globally" or similar.
-
----
-
 ## Scheduling for Other Groups
 
 When scheduling tasks for other groups, use the `target_group_jid` parameter with the group's JID from `registered_groups.json`: