Use full range for q4_0 quantization
By keeping the sign of the highest magnitude, we can make sure the
highest value maps to -8, which is currently unused.
This is a bit of a freebie since it is fully backwards compatible with
the current format.
quantize-stats output:
before(7B):
q4_0 : mse 0.00000492, maxerr 0.14257812
after(7B):
q4_0 : mse 0.00000386, maxerr 0.18200684
(Most layers have reduced maxerr under this rule, but the total max
error is indeed slightly higher)
