Get previous or next floating-point number
mps_nearfloat( op, dir )
multi-precision matrix
string "succ" or "pred"
Computes for each element of op
the next higher
(dir = "succ") or lower (dir = "pred") floating point number in the
precision of op
.