chore: use connected peers instead of discovered peers for protocols

[danisharora099]

Problem

There is a problem in the peers we choose during a protocol request: we currently get discovered peers while we should be getting and using peers that we have an open connection with.

Solution

Use connected peers instead of discovered peers that might not be connected

Notes

Interesting how we didn't run into any problems with this until now

[github-actions]

size-limit report 📦

Path	Size	Loading time (3g)	Running time (snapdragon)	Total time
Waku core	81.96 KB (0%)	1.7 s (0%)	2.4 s (+5.7% 🔺)	4 s
Waku Simple Light Node	261.34 KB (+0.03% 🔺)	5.3 s (+0.03% 🔺)	4.2 s (+44.37% 🔺)	9.4 s
ECIES encryption	31.94 KB (0%)	639 ms (0%)	1.7 s (+15.3% 🔺)	2.3 s
Symmetric encryption	31.93 KB (0%)	639 ms (0%)	1.2 s (-27.79% 🔽)	1.9 s
DNS discovery	123.17 KB (0%)	2.5 s (0%)	2.3 s (-25.52% 🔽)	4.8 s
Privacy preserving protocols	119.3 KB (-0.01% 🔽)	2.4 s (-0.01% 🔽)	1.9 s (-50.49% 🔽)	4.3 s
Light protocols	79.48 KB (+0.09% 🔺)	1.6 s (+0.09% 🔺)	2.1 s (-4.73% 🔽)	3.7 s
History retrieval protocols	78.38 KB (+0.09% 🔺)	1.6 s (+0.09% 🔺)	1.3 s (-19.38% 🔽)	2.8 s
Deterministic Message Hashing	5.92 KB (0%)	119 ms (0%)	405 ms (+151.04% 🔺)	523 ms

[adklempner]

Is there an easy way to have a discovered peer that's disconnected? Might be nice to have a test to ensure that peer isn't included

[danisharora099]

hum can you clarify what you mean?
like for a peer that doesn't support the protocol but is discovered -- it should not be connected, right?

--
merging this for now, please feel free or let me know to add a test case for that as a followup