getting ready to integrate the div/sqrt unit.

The area-optimized (i.e. speed = false) version doesn't use the PPM;
it uses behavioral A*B+C then does LZA afterwards.

x - Log2(y) is noticeably slower than a hand-rolled mux tree

Conflicts:
	chisel
	sbt/project/build.scala
	src/main/scala/anyToRecodedFloatN.scala
	src/main/scala/mulAddSubRecodedFloatN.scala
	src/main/scala/normalize64.scala
	src/main/scala/normalizeN.scala
	src/main/scala/shift_right_track_lsbs.scala
	src/main/scala/shift_round_position.scala
	src/main/scala/top.scala