Add support of more ARM VFP instructions in the assembler.
They were introduced in ARM VFPv4.

"FMULAF/FMULAD   Fm, Fn, Fd": Fd = Fd + Fn*Fm
"FNMULAF/FNMULAD Fm, Fn, Fd": Fd = -(Fd + Fn*Fm)
"FMULSF/FMULSD   Fm, Fn, Fd": Fd = Fd - Fn*Fm
"FNMULSF/FNMULSD Fm, Fn, Fd": Fd = -(Fd - Fn*Fm)

The multiplication results are not rounded.

Change-Id: Id9cc52fd8e1b9a708103cd1e514c85a9e1cb3f47
Reviewed-on: https://go-review.googlesource.com/62550Reviewed-by: Cherry Zhang <cherryyz@google.com>
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>