Initialize gradient in HessianResult to avoid undef errors when

[andreasnoack]

using BigFloats in Hessian computations

Fixes #8

I wanted to add a test but that would require ForwardDiff, i.e. something like

# hessian! + BigFloat #
# -----------------------------#

xb = big.(ones(2))
rb = HessianResult(xb)
ForwardDiff.hessian!(rb, sqrt(prod(x)), xb)
@test hessian(rb) == big.([-1 1; 1 -1]/4)

Would you like the test added to ForwardDiff?

[coveralls]

Coverage remained the same at 87.302% when pulling e4ed5ab on andreasnoack:an/big into 8d12336 on JuliaDiff:master.

[coveralls]

Coverage remained the same at 87.302% when pulling e4ed5ab on andreasnoack:an/big into 8d12336 on JuliaDiff:master.

[chriscoey]

bump

[chriscoey]

@andreasnoack do you have the ability to merge?