Catch constant solution paths

[PBrdng]

This addresses #647

The bug is still not entirely clear to me. The updates in this PR fixes the problem in #647, but by addressing constant solution paths. This cannot be the reason as outlined below.

[PBrdng]

The code that didnt work in #647 works at the current stage of this patch. However, before the patch

using HomotopyContinuation
@var x y
F = System([(y - 50) * (x - y)], variables = [y], parameters = [x])
solve(F, [50.0]; start_parameters = [-1.0], target_parameters = [-2.0])

also worked. So the reason cannot just be constant paths. The problem must be with 0 as start solution.
Notice: the following still fails:

using HomotopyContinuation
@var x y
F = System([y-x], variables = [y], parameters = [x])
solve(F, [1.0]; start_parameters = [1.0], target_parameters = [-2.0])

The reason seems to be that the WeightedNorm gives a second derivative of size 1e6 around y=0. This causes the step size to become essentially zero.