Refactor syncWithPrimary and replicaTryPartialResynchronization

[nitaicaro]

In continuation to #945

syncWithPrimary:

• Refactored returns and memory frees to single goto error/goto ok label
• Refactored repl_state state machine from if-else format to switch-case format
• Besides changing the repl_state, all state machine logic moved to helper functions

replicaTryPartialResynchronization:
This function was performing two different jobs based on the value of the read_reply argument -

• read_reply == 0: Sends the PSYNC command to the primary server.
• read_reply == 1: Reads and processes the reply to the PSYNC command.

This change simplifies the logic by clearly separating the writing and reading stages of the PSYNC process.

The PR was split into individual commits for easy review.

[codecov]

Codecov Report

Attention: Patch coverage is 86.98630% with 19 lines in your changes missing coverage. Please review.

Project coverage is 70.85%. Comparing base (9f4503c) to head (a071443).
Report is 2 commits behind head on unstable.

Files with missing lines	Patch %	Lines
src/replication.c	86.98%	19 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff              @@
##           unstable    #1476      +/-   ##
============================================
- Coverage     70.86%   70.85%   -0.02%     
============================================
  Files           119      119              
  Lines         64859    64882      +23     
============================================
+ Hits          45963    45972       +9     
- Misses        18896    18910      +14     

Files with missing lines	Coverage Δ
src/replication.c	87.60% <86.98%> (+0.07%)	⬆️

... and 11 files with indirect coverage changes

[PingXie]

This kind of refactoring is really valuable for the long-term health of Valkey. It makes the codebase easier to understand, maintain, and improve, which ultimately helps us deliver better features faster. Thanks @nitaicaro and @naglera!

My main focus in this review is on the high level code structure, which looks great to me. That said, the use of goto in syncWithPrimary stands out. Let's aim to eliminate all goto statements from that function. I'll take another look and provide more detailed feedback once that's addressed.

[PingXie]

nit - I don't remember we start the comment block with /**?

Suggested change

	/**
	* Handles the initial step of the partial resynchronization process by
	/* Handles the initial step of the partial resynchronization process by

[PingXie]

nit - I believe this is more common.

Suggested change

	* step is to call replicaProcessPsyncReply().
	*/
	* step is to call replicaProcessPsyncReply(). */

[PingXie]

I personally like using double empty lines between functions but the convention is single empty link I think

[PingXie]

These goto labels (no_response_error, error, write_error, and ok) have really gone too far now. Let's consolidate them into a single syncWithPrimaryHandleError function. I am hopeful that we can remove all the gotos in this function.