bb fall-through onto unreadable page handling is not 100% transparent

A bb that falls through onto a subsequent unreadable page is not handled perfectly.

check_new_page_contig() won't complain, b/c It decodes first and throws away the new instr if it's off onto an incompatible vmarea: so we'll fault before we call it.

We directly generate a fault to the app in our signal handler, which is not transparent if the app were to catch and handle the execution fault: we never executed the intervening instrs in the bb that we successfully decoded.

We don't really want to pay the cost of the setjmp for a try around every decode, so it's not clear the simplest way to solve this.