memory usage in soter_asym_cipher_init()

[outspace]

I'm look at soter_asym_cipher_init() in BoringSSL and OpenSSL threads. In bad case of macro we call EVP_PKEY_CTX_free than we call EVP_PKEY_CTX_free again in soter_asym_cipher_cleanup that's why it shoud be nulled in BoringSSL too.

[Lagovas]

agree, we should nullify freed pointers to avoid problems with second free call

[ilammy]

Resource deallocation has been improved. Stored pointers in key generations routines are now zeroed out after freeing them.