After EDIT in Question:
The operations in step 1, 2 and 3, you mentioned, are actually
O(1) it is not meant that it is going to take 1 step only but it means that whatever time it is taking is CONSTANT. That constant can be small or large, but as you increase number of nodes in your tree, these basic operations are not going to take more time. Be it on a tree with 3 nodes or 200 nodes, one of these operation is going to take the same amount of time each time.
For an example, take operation 1
Compare 2 children. It is always going to take same number of comparison(here 1).
But if you say the sequence of step 1, 2, 3, 4 in order, as an algorithm. Then it is not
O(1) and instead
O(log n) with n being the number of nodes in the tree. Reason being, the operation 1-3 are done from a bottom pair of two nodes, propagating to top root node. So, in worst case it will traverse equal to height of the tree which is