[core]Fixing the group option that should be taken from a socket

[ethouris]

Changed: Added a possibility to return the option value that was set on the socket, if there is at least one member.

The problem: this should rather concern only selected types of options, not all options. So far probably it should only
apply to the following options:

• SRTO_RCVLATENCY, SRTO_PEERLATENCY, SRTO_LATENCY
• SRTO_CRYPTOMODE

Without this the test will fail, and it's also not expected. We need this to work only for those options that could be altered as a result of the connection and involved parameter exchange.

Addresses #2918.

[maxsharabayko]

For SRTO_KMSTATE I suggest returning the least secure state among member sockets.
The order of security strength should be the following (from least secure to the most secure):

1. SRT_KM_S_BADCRYPTOMODE – it is a connection rejection code, so maybe should be ignored or set only if no other members are there? 🤔
2. SRT_KM_S_BADSECRET – the password is wrong (set differently on each party); encrypted payloads won't be decrypted.
3. SRT_KM_S_NOSECRET – the sending party has set a passphrase, but the peer did not. In this case the sending party can receive unencrypted packets from the peer, but packets it sends to the peer will be encrypted and the peer will not be able to decrypt them.
4. SRT_KM_S_SECURING – pending security (HSv4 only).
5. SRT_KM_S_UNSECURED – no encryption/decryption.
6. SRT_KM_S_SECURED

[maxsharabayko]

For SRTO_PBKEYLEN I suggest returning the shortest key length among member sockets.