Clarify Bonacich centrality discussion

[stevelinberg]

Again, the language is too ambiguous and is guessing too much. The R documentation for igraph::power_centrality() says that it was "copied" from sna::bonpow(), but the source for the two functions is not the same, and it returns at least one simply-visible difference in the present code (China, P.R.: Mainland).

The discussion of singular matrices seems out of place, and the suggestion to just try the other library if one produces NaN needs more exposition.