Actually, I found the documentation reasonably poor. It does a good job of explaining the API but does little to provide a would-be user with any insights into why the search works, how it works or how efficient it is. It doesn't even discuss why (or when) I might choose one of the three given heuristic functions over the others.
i.e. despite its length, it's not actually a very useful document.
While I agree that what you listed would be useful to include I think it's important to note that it is documentation of his implementation (what the various parameters are for) and not of how A* works. To that end he did a good job. I especially like the little ideas he throws in under "Extending astar.js." He links to the Wikipedia entry for those who want to know more about how A* works.
Suppose the C++ STL documentation, such as it is, linked you to CLRS. Would you regard it as adequate or would you demand more? I don't expect a rigorous academic treatment here but at least something to tell me (a) how efficient the implementation is and (b) why (or when) I might want to use X or Y provided feature over Z. For example: why is the Euclidean heuristic "slow"? Is it a shitty implementation? Is it because it's underestimating the goal distance too much? Another gripe: the author doesn't even tell me what kind of open and closed list implementations are in use!
I don't mean to be a hater but this fails at contributing something novel (which it doesn't), it fails as a learning resource which teaches something interesting (which it doesn't) and it fails as a well documented programming hack (which it isn't).
i.e. despite its length, it's not actually a very useful document.