tls: make tls a little bit faster

[indutny]

Compile OpenSSL with inline assembly for big numbers.

Benchmarks:

Original node

Concurrency Level:      100
Time taken for tests:   25.138 seconds
Complete requests:      10000
Failed requests:        0
Write errors:           0
Total transferred:      800000 bytes
HTML transferred:       50000 bytes
Requests per second:    397.81 [#/sec] (mean)
Time per request:       251.375 [ms] (mean)
Time per request:       2.514 [ms] (mean, across all concurrent requests)
Transfer rate:          31.08 [Kbytes/sec] received

Patched node:

Concurrency Level:      100
Time taken for tests:   18.616 seconds
Complete requests:      10000
Failed requests:        0
Write errors:           0
Total transferred:      800000 bytes
HTML transferred:       50000 bytes
Requests per second:    537.17 [#/sec] (mean)
Time per request:       186.161 [ms] (mean)
Time per request:       1.862 [ms] (mean, across all concurrent requests)
Transfer rate:          41.97 [Kbytes/sec] received

[TooTallNate]

Will this work on ARM?

[indutny]

No, it even won't work on x86

[bnoordhuis]

Easy win is easy. LGTM and nice work, Fedor. Land it in master for now.

[mranney]

Woah, awesome. More speedups like this, please.

[indutny]

@bnoordhuis @isaacs are we going to merge this into 0.8?

[indutny]

Otherwise, landed there d923269

[bnoordhuis]

are we going to merge this into 0.8?

Possibly. Let's use master as a testing ground.

[indutny]

Another patch for probable landing in 0.8: f210530