Adjust vid.width floating point comparson for CHAT
-Wfloat-equals compile errors that Alam enabled for the
next branch. See STJr/SRB2!383 (merged) for that discussion.
!= == comparisons are invalid for floating point due to precision weirdnesses -- decimals in the 6th or 10th place may be off by 1.
So to do a not-equal comparison, you subtract the two values and see if that difference is non-zero.
fabsf(X-Y) > 1.0E-36f checks for non-zero
X-Y by seeing if the abs value of that difference is greater than
.00000...000001 to the 36th decimal place.