Support for LogExpFunctions && better errors

Hey,

import TaylorDiff
import LogExpFunctions
t = TaylorDiff.make_seed(1.2, 1.0, Val{3}())
log1p(exp(t))
LogExpFunctions.log1pexp(t)

yields:

julia> import TaylorDiff

julia> import LogExpFunctions

julia> t = TaylorDiff.make_seed(1.2, 1.0, Val{3}())
TaylorScalar{Float64, 3}(1.2, (1.0, 0.0, 0.0))

julia> log1p(exp(t))
TaylorScalar{Float64, 3}(1.4632824673380311, (0.7685247834990176, 0.08894722032340283, -0.015923022053454108))

julia> LogExpFunctions.log1pexp(t)
1.4632824673380311

julia> 

Two questions:

1. How hard would it be to support LogExpFunctions
2. Why, why and why dont we error in those cases ? Took me ages to understand what was going on...

Start by reproducing the shown TaylorDiff.make_seed example with log1p(exp(t)) and LogExpFunctions.log1pexp(t), then inspect the relevant TaylorDiff entry points for these operations and their error behavior. Done means LogExpFunctions.log1pexp works with the TaylorScalar example and unsupported cases produce informative errors.