test envelope example for n > 1 (check vs matlab)

[chriscoey]

and timing comparisons into #23

[chriscoey]

this was broken by #34 because I incorrectly used the same primitive cone object multiple times in the problem data construction. that's fixed, and I added tests for dense vs sparse A and n > 2.

however, while the tests run fine for n <= 2, at n = 3 I am consistently getting the problem coming back nearly infeasible.

this suggests that there may be a problem with the clenshaw-curtis weights calculations for approximate fekete points. these values not used in the namedpoly tests so they aren't tested there. debugging is WIP.