refactor(redis): migrate remaining get_redis_client(async_client=True) callers to get_async_redis_client #4059

Problem

PR #4051 introduced get_async_redis_client() as a structurally-safe replacement for the error-prone await get_redis_client(async_client=True, ...) pattern, but only migrated one caller (knowledge/base.py). There are ~20+ remaining callers across the backend still using the old pattern, including:

• events/stream_manager.py
• memory/working_memory.py
• plugin_manager.py
• enhanced_orchestration/orchestrator.py
• utils/cache_manager.py
• api/workflow_state.py
• And others

These callers are susceptible to the original #3962 bug if the await is accidentally removed in a future edit.

Fix

Grep for get_redis_client(async_client=True across the backend and migrate each call site to get_async_redis_client(database=...).

Discovered

During code review of PR #4051 (fix for #3962).