Release GIL while doing curve functions.

[mayfield]

Releasing the GIL during compute intensive functions lets threaded
applications scale more effectively. Simple benchmarks on a quad core i7
show a speed up of ~573%.

Simple benchmark used:

import axolotl_curve25519 as curve
import os
import threading

randm32 = os.urandom(32)
randm64 = os.urandom(64)

private_key = curve.generatePrivateKey(randm32)
public_key = message = curve.generatePublicKey(private_key)
import time

calcs = 0

def some():
    for i in range(10000):
        curve.calculateAgreement(private_key, public_key)
        signature = curve.calculateSignature(randm64, private_key, message)
        curve.verifySignature(public_key, message, signature)
        global calcs
        calcs += 1

for test in range(3):
    print("\nTEST:", test)
    calcs = 0
    threads = [threading.Thread(target=some) for i in range(4)]
    for x in threads:
        x.start()
    s = time.time()
    for x in threads:
        x.join()
    took = time.time() - s
    print("Took", took, calcs / took)

[mayfield]

Forgot to mention, single threaded benchmarks didn't show any measurable slowdown.