Incorrect limit evaluation on Chaikin regression shapes

[manuelkNVDA]

Both Catmark & Loop show deltas:

catmark:

loop:

[manuelkNVDA]

Inspection of the surface neighborhood does not reveal any obvious issues - crease tags appear correctly formed

Neighborhood hash = 0x63E30C6F edge=0 {
        face vert counts = (5) { 4, 4, 4, 4, 4 }
        face verts = (20) { 0, 1, 2, 3, 0, 3, 4, 5, 0, 5, 6, 1, 1, 6, 7, 2, 2, 7, 4, 3 }
        corner verts = (0) {  }
        corner sharpness = (0) {  }
        crease verts = (16) { 0, 1, 0, 3, 0, 5, 1, 2, 1, 6, 2, 3, 2, 7, 3, 4 }
        crease sharpness = (8) { 5.000, 5.000, 1.000, 5.000, 1.000, 5.000, 1.000, 1.000 }
        control points = (20) { 22, 23, 1, 0, 22, 0, 2, 20, 22, 20, 21, 23, 23, 21, 3, 1, 1, 3, 2, 0 }
        remap table = (8) { 22, 23, 1, 0, 2, 20, 21, 3 }
}

note: Bfr regression test appear to be showing the exact same problem, so this may not be a Tmr bug after all