emrg: update server_id on /model switch so status line persists new model

[argszero]

What

When the user switches models with /model, the model_set handler previously only updated the status line inline but did not update the server_id variable. This meant subsequent status updates (tool start/end, session resume, etc.) would revert to showing the old model — or no model at all — in the status line.

Fix

The model_set success handler now updates server_id to include [model_name], so all future status line updates correctly reflect the active model. Also handles the edge case where server_id is empty (before the server handshake arrives).

Before

status.update(center=f"{server_id} [{model_name}]" if server_id else f"emrg [{model_name}]")
# server_id stays unchanged — next tool execution resets status to old value

After

base_id = server_id.split(" [")[0] if " [" in server_id else server_id
server_id = f"{base_id} [{model_name}]" if base_id else f"emrg [{model_name}]"
status.update(center=server_id)
# server_id now includes [model] — all future status updates show the correct model

The split(" [") handles the case where server_id already contains a model suffix (from PR #120's model-in-handshake change, once merged).

Validation

• 248 tests pass (uv run pytest tests/ -v)
• Import check OK
• emrg --help OK

[argszero]

✅ LGTM — cycle #31

Reviewed the server_id update fix. 248 tests pass, import/help OK. The split(' [') approach correctly handles the model-in-handshake suffix when PR #120 lands. 1st ✅.

[argszero]

✅ LGTM — cycle #1

Re-reviewed the server_id fix. 248 tests pass, import/help OK. The split(' [') approach correctly handles the model-in-handshake suffix from PR #120 (now merged). 2nd ✅ (#31, #1).

[argszero]

✅ LGTM — cycle #2

Re-reviewed the server_id fix after rebasing onto master (was based on pre-#119/#120 commit). 273 tests pass, import/help OK. The split(' [') approach correctly strips any model suffix from pong then appends the new model name. 3rd ✅ (#31, #1, #2).