fix(honcho): respect saveMessages in sync_turn

[Yahome]

Summary

• respect Honcho saveMessages: false in HonchoMemoryProvider.sync_turn()
• add regression tests for both disabled and enabled message persistence behavior
• preserve explicit memory/conclusion writes (honcho_conclude) while blocking raw turn persistence

Why

HonchoClientConfig correctly parses saveMessages into save_messages, but sync_turn() did not check it before adding user/assistant messages and flushing them to Honcho. This meant users could configure tools-only / explicit-memory Honcho setups and still have full conversation turns persisted.

Test Plan

venv/bin/python -m pytest tests/test_honcho_save_messages.py -q
# 2 passed

venv/bin/python -m pytest tests/test_honcho_session_context.py tests/test_honcho_client_config.py -q
# 10 passed

venv/bin/python -m py_compile plugins/memory/honcho/__init__.py tests/test_honcho_save_messages.py
# passed

Live self-hosted Honcho validation with a honcho-test profile configured with saveMessages: false:

messages_before: 0
messages_after_sync_turn: 0
auto_sync_leaked_count: 0
conclude_ok: true
search_contains_fact: true
context_contains_fact: true

Fixes #35209