Expand ECDH test vectors

[marsella]

Right now we have the CAVP test vectors to demonstrate the correct behavior for ECDH with P-256. We can expand this in two ways. First, we don't have any negative test vectors (e.g. to show that things aren't trivially correct / even invalid keys are approved). It might be good to think about whether there's a principled way to do this or if we should just mash some of the valid vectors to get invalid ones. Second, we could instantiate ECDH for more curves. NIST SP 800-186 allows all the prime-field curves to be used for EC key agreement (section 3.1.2, Table 2), so we might instantiate all of those and pull in all the corresponding CAVP test vectors.

• Add some negative test cases for each method in ecdh
• Add additional P-curves (P-224, P-384, P521) and CAVP test vectors