LibCrypto+LibWeb: Support P-521 with OpenSSL

[devgianlu]

This PR introduces P-521 support by rewriting the SECPxxxr1 implementation using OpenSSL, this should be the last remaining piece of implementation for WebCryptoAPI. After this I expect us to pass all WPT tests except ~100.

[alimpfard]

I took a brief look, the only issue I noticed was addressed in the last commit so 👍
Will do a more thorough review in a bit.

[alimpfard]

Ah yes,

102/161 Test #102: TestCurves ...........................***Failed    0.09 sec
Running test 'test_x25519'.
Completed test 'test_x25519' in 3ms
Running test 'test_x448'.
Completed test 'test_x448' in 14ms
Running test 'test_secp256r1'.
Completed test 'test_secp256r1' in 0ms
Running test 'test_secp384r1'.
Completed test 'test_secp384r1' in 7ms
Finished 4 tests and 0 benchmarks in 26ms (24ms tests, 0ms benchmarks, 2ms other).
All 4 tests passed.

[devgianlu]

Ah yes,

102/161 Test #102: TestCurves ...........................***Failed    0.09 sec
Running test 'test_x25519'.
Completed test 'test_x25519' in 3ms
Running test 'test_x448'.
Completed test 'test_x448' in 14ms
Running test 'test_secp256r1'.
Completed test 'test_secp256r1' in 0ms
Running test 'test_secp384r1'.
Completed test 'test_secp384r1' in 7ms
Finished 4 tests and 0 benchmarks in 26ms (24ms tests, 0ms benchmarks, 2ms other).
All 4 tests passed.

I love it when it's so descriptive lol. Anyway, I think it's a memory leak problem again, I am building with sanitizers locally.

[alimpfard]

Code looks largely OK, there's some signed len functions that could maybe have better names for the return values though:

[devgianlu]

I love it when it's so descriptive lol. Anyway, I think it's a memory leak problem again, I am building with sanitizers locally.

Indeed, I was not freeing instances of EC_group.