Improve operations count to reflect real world DOM mutations

[WebReflection]

As discussed in https://github.com/luwes/js-diff-benchmark/issues/7#issuecomment-615875584, the way we are counting operations does not reflect the real-wolrd amount of DOM mutations triggered through the nodes:

• appendChild is 1 up to 2 mutations, if the node was already live (moved)
• insertBefore is 1 up to 2 mutations, if the node was already live (moved)
• removeChild is always only 1 mutation
• replaceChild is 2, up to 3 mutations if the node was already live (moved)

Accordingly, the counting should change too, specially the swap and the shuffle, as most libraries technically trigger similar amount of observable mutations on the tree:

• the swap consist in 4 mutations, 2 disconnect followed by 2 connect
• the shuffle consist in either removeChild followed by an insertBefore, which sum is 2, or replaceChild which counts as either 2 or 3

The table better reasons, and explain, why udomdiff performs similarly, if not faster, than other libraries when it comes to the shuffling example

[WebReflection]

P.S. the proof for mutations happening even if a node doesn't leave the tree is here https://github.com/luwes/js-diff-benchmark/issues/7#issuecomment-615652697

[WebReflection]

Following the table using the before element:

[WebReflection]

feel free to merge too, I added you as a collaborator.

I know, but I'm used to have other people eyes on my PR when it's more than just me on a repo 😉