Fix determination of staked QUIC connections

[pgarg66]

Problem

The usage of ConnectionPeerType to determine if a connection is staked is incorrect. The peer_type is an attribute of the connection_table where a QUIC connection gets stored. We maintain two tables, one for staked connections, and the other for unstaked connections. But if the staked connection is full, the new connections get stored in the unstaked connection table, even if the peer is a staked client. So ConnectionPeerType can incorrectly mark a staked connection as unstaked.

Peer's stake correctly identifies if the connection is from a staked client or not.

Summary of Changes

Update code to use peer's stake to determine if the connection is from a staked client.
Remove ConnectionPeerType and its usage.

Fixes #

[codecov]

Codecov Report

Attention: 19 lines in your changes are missing coverage. Please review.

Comparison is base (22fcffe) 81.8% compared to head (8fb5a01) 81.8%.
Report is 9 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff            @@
##           master   #34760     +/-   ##
=========================================
- Coverage    81.8%    81.8%   -0.1%     
=========================================
  Files         824      824             
  Lines      223051   223049      -2     
=========================================
- Hits       182474   182465      -9     
- Misses      40577    40584      +7     

[lijunwangs]

We can reduce some of this change by defining:

let staked_stream = params.stake > 0;

[pgarg66]

Isn't it better to have this check than declaring a new stack variable? The change is only three lines. So review should be simple.

[t-nelson]

i think i'd have preferred resolving the enum once and using that throughout. the Staked variant just needs to be Staked { lamports: u64 }

[pgarg66]

i think i'd have preferred resolving the enum once and using that throughout. the Staked variant just needs to be Staked { lamports: u64 }

Yes, that's more idiomatic. Will update the PR.

[pgarg66]

i think i'd have preferred resolving the enum once and using that throughout. the Staked variant just needs to be Staked { lamports: u64 }

Done

[lijunwangs]

lgtm