Explore use of STACK_OF instead of hashtable

[npajkovsky]

#1726 (comment)

Can you please attach the code and the steps on how to reproduce it? If you want to test pure
EVP__fetch, there is PR for perftools openssl/perftools#80

@bob-beck

[npajkovsky]

The -F enables frozen cache

STACK_OF:

$ ./evp_fetch -f MD:SHA2-256 64 -F
Average time per fetch call: 1.148093us
$ ./evp_fetch -f MD:SHA2-256 64
Average time per fetch call: 24.442072us

$ ./evp_fetch -f  CIPHER:AES-128-GCM 64 -F
Average time per fetch call: 1.252560us
$ ./evp_fetch -f  CIPHER:AES-128-GCM 64
Average time per fetch call: 24.289483us

Hashtable:

./evp_fetch -f MD:SHA2-256 64 -F
Average time per fetch call: 1.197693us
$ ./evp_fetch -f MD:SHA2-256 64
Average time per fetch call: 24.295191us

$ ./evp_fetch -f  CIPHER:AES-128-GCM 64 -F
Average time per fetch call: 1.244238us
$ ./evp_fetch -f  CIPHER:AES-128-GCM 64
Average time per fetch call: 23.303556us

Hashtable flamegraph

STACK_OF flamegraph

Baseline

The performance difference between STACK_OF and HT is less than 1us, hence unmeasurable on my M4 Pro.