Add NormalError cost function

[sarlinpe]

Mahalanobis distance between two parameters.

Example:

x1 = np.zeros(3)
x2 = np.ones(3)
x1_prior = np.arange(3)
covar = np.eye(3)

prob = pyceres.Problem()
loss = pyceres.TrivialLoss()
prob.add_residual_block(pyceres.factors.NormalPrior(x1_prior, covar), loss, [x1])
prob.add_residual_block(pyceres.factors.NormalError(covar), loss, [x1, x2])

options = pyceres.SolverOptions()
summary = pyceres.SolverSummary()
pyceres.solve(options, prob, summary)
np.testing.assert_allclose(x1, x1_prior, atol=1e-7)
np.testing.assert_allclose(x2, x1, atol=1e-7)

[B1ueber2y]

Thanks. I confirm that this feature works in my test.