Add handshaking for Raft CLUSTER MEET

[murphyjacob4]

Building off of #3530 - adds a simple handshake request/response between nodes.

Does not perform any Raft group mutations - the nodes are tracked as RAFT_ROLE_NON_MEMBER and are subject to removal by timeout.

The expectation is that nodes will decide individually who will join each other based on a somewhat deterministic mechanism to be implemented in a future PR.

[codecov]

Codecov Report

❌ Patch coverage is 70.60703% with 92 lines in your changes missing coverage. Please review.
✅ Project coverage is 76.51%. Comparing base (6b93134) to head (daba9d1).

Files with missing lines	Patch %	Lines
src/cluster_raft.c	67.94%	92 Missing ⚠️

Additional details and impacted files

@@              Coverage Diff               @@
##           cluster-v2    #3532      +/-   ##
==============================================
- Coverage       76.59%   76.51%   -0.09%     
==============================================
  Files             160      161       +1     
  Lines           79288    79582     +294     
==============================================
+ Hits            60733    60890     +157     
- Misses          18555    18692     +137     

Files with missing lines	Coverage Δ
src/cluster.c	90.66% <100.00%> (-0.12%)	⬇️
src/cluster_legacy.c	91.24% <100.00%> (-0.51%)	⬇️
src/cluster_state.c	88.74% <100.00%> (+0.63%)	⬆️
src/config.c	78.33% <ø> (ø)
src/server.c	89.45% <100.00%> (-0.11%)	⬇️
src/server.h	100.00% <ø> (ø)
src/cluster_raft.c	67.94% <67.94%> (ø)

... and 22 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[zuiderkwast]

I think we can keep both our prototypes alive for a while to discover parth forward and compare them. I think we can wait with merging this to cluster-v2.

Yours has the key-value style Raft with CAS, binary protocol. Mine has domain-specific entry types, text protocol. Both probably have pros and cons. I don't feel strongly about the text/binary protocol nor the CAS style. I picked what seemed to be the simplest possible protocol to start with because I wanted to move forward to see how much of the existing cluster tests I could get passing. I believe the learnings from making all existing things work (automatic failover with ranks, manual failover with pause writes, etc) has value regardless, and I think it'd be not that hard to change the wire protocol later.

However, clusterDelSlot needs the refactoring from this PR. I'd like to merge it. It should be strait-forward to extract it from this PR. I can extract it and open it as a separate PR...

[zuiderkwast]

I notice this missing piece too.

cluster_link.c calls clusterDelNode which is defined in cluster_legacy.c but declared in cluster_state.h. It's incomplete work from the cluster protocol separation refactoring.

I see you moved clusterDelNode to cluster_state.c and made it call this callback.

It'd be good to have this as a separate commit or PR.

[zuiderkwast]

I've extracted this part to a separate PR now: #3542