Consolidate various BLOCKED_WAIT* states

[PingXie]

There are currently three block types: BLOCKED_WAIT, BLOCKED_WAITAOF, and BLOCKED_WAIT_PREREPL, used to block clients executing WAIT, WAITAOF, and CLUSTER SETSLOT, respectively. They share the same workflow: the client is blocked until replication to the expected number of replicas completes. However, they provide different responses depending on the commands involved. Using distinct block types leads to code duplication and reduced readability. This PR consolidates the three types into a single WAIT type, differentiating them using the pending command to ensure the appropriate response is returned.

Fix #427

[codecov]

Codecov Report

Attention: Patch coverage is 95.23810% with 1 lines in your changes are missing coverage. Please review.

Project coverage is 70.04%. Comparing base (168da8b) to head (a0cefda).
Report is 3 commits behind head on unstable.

Additional details and impacted files

@@             Coverage Diff              @@
##           unstable     #562      +/-   ##
============================================
- Coverage     70.22%   70.04%   -0.19%     
============================================
  Files           109      109              
  Lines         59956    59955       -1     
============================================
- Hits          42104    41993     -111     
- Misses        17852    17962     +110     

Files	Coverage Δ
src/cluster_legacy.c	86.36% <100.00%> (-0.10%)	⬇️
src/networking.c	85.35% <100.00%> (ø)
src/replication.c	87.20% <100.00%> (+0.04%)	⬆️
src/blocked.c	91.45% <92.85%> (-0.42%)	⬇️

... and 6 files with indirect coverage changes

[zuiderkwast]

Nice. Please add some useful text in the description, rather than just a link to the issue. We always ask that from contributors. It's helpful when reviewing, it's used as commit message when merging and it's good for reference later, even if the PR is not material for release notes.

[PingXie]

Will do

[zuiderkwast]

Very good cleanup. The PR description helped me a lot! Thanks.

If you decide to take my suggestions, don't forget to add signoff.

[PingXie]

If you decide to take my suggestions, don't forget to add signoff.

I feel that comparing to the cmd structure point is cleaner/more readable but yes it is very subjective.

[zuiderkwast]

If you decide to take my suggestions, don't forget to add signoff.

I feel that comparing to the cmd structure point is cleaner/more readable but yes it is very subjective.

Pros and cons. It's perhaps cleaner, but it's more important IMO to be consistent with other code doing the same thing. (The alternative is to change all the existing command comparisons to your style.)

Avoiding the need to initialize and store these extra pointers is another point.

[PingXie]

but it's more important IMO to be consistent with other code doing the same thing.

You got me on "consistency" :-). I will switch to proc.

[zuiderkwast]

Looks good, but there's some crash in test "XREAD: XADD + DEL + LPUSH should not awake client".

[PingXie]

This is not a crash but a behavior change caught by the test. The unification of the three blocked types,, incorrectly, allows an LPUSH operation to wake up a stream reader.

On a closer look, I don't think there is an easy way to unify these 3 states, for two reasons:

1. each wait (list/zset/stream) can be initiated by multiple commands. it is quite a hassle to track down all these commands;
2. stream can be waited on even when it is non-existent, which makes it impossible to infer the data type from blockForKeys implicitly

I reverted the last commit.