Fix scratchpad double reap and redundant parent operations

[jbms]

This commit fixes a double-reap and redundant operations when sending a tiled container to the scratchpad.
When the container is tiled, root_scratchpad_add_container calls container_set_floating, which detaches the container from its tiled parent and reaps the parent if it becomes empty.
However, root_scratchpad_add_container was then:

1. Calling container_reap_empty on the parent again, causing a redundant/double-reap.
2. Calling arrange_container(parent) on the parent even if it has already been marked for destruction and is about to be reaped.

We fix this by:

1. Removing the redundant container_reap_empty(parent) call, as the parent is already reaped in container_set_floating (if it was tiled) or it is NULL (if it was already floating).
2. Checking the parent->node.destroying flag directly to verify if the parent container is being destroyed. If it is being destroyed, we set parent to NULL to avoid arranging it, falling back to arranging the workspace instead.

[smlx]

Removing the redundant container_reap_empty(parent) call, as the parent is already reaped in container_set_floating (if it was tiled) or it is NULL (if it was already floating).

I don't see that change in the diff. Is this perhaps an outdated description?