Double boundary with custom boundary

[abhranildas]

When a custom boundary is supplied, the boundary drawn will often be double. The boundary computed with dist a and that computed with dist b (just negative of the coefficients) won't exactly line up, like they do when no custom boundary is supplied and it computes the optimal bdry with dist a, and with dist b (which ought to also just be a negation).

[abhranildas]

When a custom boundary is supplied, there's an additional calculation, of transforming those coefficients, which doesn't occur when the optimal boundary is requested. There's probably a numerical inaccuracy in this calculation so that slightly different results are returned when this transformation is done wrt dist a vs wrt dist b.