Added seeder API. Also overhauled compile-time detection of features.
Fixed IV processing for CBC decryption with AES (x86ni implementation) when data length is not multiple of 64 bytes (the bug was breaking TLS 1.0 AES/CBC on recent x86 systems).
Fixed compilation for GCC 4.4 to 4.8 (AES-NI opcodes; intrinsics headers require target options to be set).
Added AES+GHASH implementation using AES-NI opcodes; also ARM-Thumb assembly for faster Montgomery multiplication on Cortex-M0+. Added selection functions for "default" implementations.