Fix error of __list_find

[RinHizakura]

The node on the linked list should follow the ordering by its key. However, the function list_insert doesn't actually do this because of the wrong implementation of __list_find.

There are three main fixes in this patch. First, once we should do synchronization, we have to iterate the list again but not return from the function. Second, the while loop should iterate until the last node but not the second last one. Finally,
under the normal situation that the tail node with key UINTPTR_MAX should not be deleted (unless the end of the program), some codes are never executed which can be removed.

[jserv]

I would expect the following comments:

On a CAS failure, the search function, "__list_find", will simply have to
go backwards in the list until an unmarked element is found from which
he search in increasing key order can be started.

For the implementation to be correct, the predecessor pointers only
have to fulfill that any item always has a path back to the head of the
list. When a new item is inserted into the list, the predecessor pointer
of the successor is set with an atomic store to point to the new element.
Likewise, when an item is removed, the predecessor pointer of the
successor item is updated to skip the removed item. This suffices to
maintain the invariant, but through long sequences of concurrent
insertions and deletions, backwards pointers can become imprecise
in that they skip many items that are actually in the list.

My implementation was a functional rewriting of paper A more Pragmatic Implementation of the Lock-free, Ordered, Linked List. You can check its implementations and benchmark programs: lockfree-linked-list.

[RinHizakura]

Thanks for the information. I changed the comment according to yours.