mesh: Multi-thread safety issue

I found a multi-thread safety issue in mesh stack code and it still exist in the newest version.
When "seg_ack" was called by callout module, there was anthoer thread which receive mesh datas was calling "seg_rx_reset()" and set the "rx->sub" to NULL, then "seg_ack" called "send_ack" and acess the rx->sub->net_idx, so it crashed.
The adv-sending task and ble-data-receiving task and each "callout" are not in the same thread, why don't you add the Multi-thread protection mechanism in mesh stack? or something I used the wrong way?
Waiting for your replies! thank you so much!

Start by tracing the mesh-stack paths named in the report: seg_ack, send_ack, seg_rx_reset(), and the callout, advertising, and BLE data-receiving tasks. Reproduce the concurrent reset and acknowledgement scenario, then verify that seg_ack cannot access rx->sub->net_idx after seg_rx_reset() clears it, without introducing a crash or race.

Written by the indexing model from the issue text.