mgmt: smp: Fix race condition by using same work queue

Fix possible race condition where SMP client might
release the network buffer before system worker queue
has processed it.

In smp_client uses shared resources like worker queue
linked list and network buffers without maintaining any
thread safety.

For unknown reasons, retry timeout handling is pushed
into system worker queue while the actual transmission
is handled from SMP work queue.

Fix the issue by using the same SMP work queue
for both delayable timeout handling as well as
transmission handling.

Signed-off-by: Seppo Takalo <seppo.takalo@nordicsemi.no>