fix: downgrade 4 TODOs to XXX per repo standard

No vendor files changed. Classification of non-vendor TODOs:

TODO → XXX (someday-maybe, no commitment):
  packages/bash-local/src/run.ts:248
    — XXX(stateful-shell): design reference for future workflows;
      current spawn-per-call is deliberate and works fine.

  packages/tool-bash/src/index.ts:23,165
  packages/tool-bash/tests/tools.spec.ts:413
    — XXX(tool-bash-owner-hmr): HMR-only issue; dev-only,
      single-user cooperative editor, not a trust boundary.

All other TODOs kept as-is:
  TODO(demo)           — should fix for production deployment
  TODO(sub-agents)     — planned feature
  TODO(review)         — validation pending real adapters
  TODO(http)           — should refactor raw fetch
  TODO(permissions)    — important security feature
  TODO(rfc010-*)       — deferred ACP features, should land when resources permit
  parallel execution   — phase 1 sequential, performance improvement

packages/tool-bash/tests/tools.spec.ts

@@ -410,7 +410,7 @@ describe('background task ownership (cross-session isolation)', () => {
   })
 
   it('documents the HMR caveat: an independent tool-bash reload resets ownership', async () => {
-    // The ownership map is per-plugin-instance (TODO(tool-bash-owner-hmr)). When
+    // The ownership map is per-plugin-instance (XXX(tool-bash-owner-hmr)). When
     // ONLY tool-bash is reloaded (bash/executor + task survive), the new instance
     // has an empty map, so the previously-owned task becomes unowned (open). This
     // test pins that documented behavior — a regression here (e.g. an accidental