gravwonbev: vertical polygon sides are counted twice

[Advik-B]

In gravwonbev, the branch for a vertical side (grav_2dpolybodies.jl line 409) returns

g = factor*x1*lor21

If you take the general expression as x21 goes to 0 (so x1 = x2, x1*z2 - x2*z1 = x1*z21 and R = z21^2), the θdiff term drops out and you're left with 0.5*factor*x1*lor21. So every exactly vertical side counts double.

My guess is the 0.5 got lost when this was adapted from the magnetic version, since in tmagwonbev it's already inside lor21 = 0.5*(log(r2)-log(r1)) while in gravwonbev it's lor21 = log(r2) - log(r1) (line 376).

Repro on b233034:

using MagGravPoly.MG2D

xz = [3.0 0.0]

rect = [-10.0 30.0; 10.0 30.0; 10.0 10.0; -10.0 10.0]
body = GravPolygBodies2D([[1, 2, 3, 4]], rect, [400.0]; ylatext=nothing)
println("talwani: ", tgravpolybodies2Dgen(xz, body, "talwani"))
println("wonbev:  ", tgravpolybodies2Dgen(xz, body, "wonbev"))

# same box with two corners nudged by 1e-9 so neither side is exactly vertical
rect2 = [-10.0 30.0; 10.0 + 1e-9 30.0; 10.0 10.0; -10.0 + 1e-9 10.0]
body2 = GravPolygBodies2D([[1, 2, 3, 4]], rect2, [400.0]; ylatext=nothing)
println("wonbev, nudged: ", tgravpolybodies2Dgen(xz, body2, "wonbev"))

talwani: [0.10323558366559354]
wonbev:  [0.1857224974488919]
wonbev, nudged: [0.10323558366580585]

So a plain rectangle comes out about 1.8x too big, and 1e-9 away from vertical it's fine. g = 0.5*factor*x1*lor21 should fix it. The magnetic x21 == 0 branch agrees with talwani on the same box, so as far as I can tell it's only gravity.