Fix cycle counting for relative branches. The original creator of the cycle counting code appears to have mangled two different approaches to counting the cycles for relative branches into the final product which led to counting too many cycles for the branch instructions. The comments in the cycle table were correct but the code was doing some incorrect things. Also add the missing LBRN entry to the cycle table. The one labelled LBRN was not, in fact, LBRN, but one of the actual useful branch instructions.