tools: bugfix block-generator to handle conduit's Init block requests

[tzaffi]

Summary

Conduit's recent PR: algorand/conduit#89 broke the assumption that block-generator was making that all block requests must advance the ledger. This PR addresses this breakage by allowing more leniency to a caller which requests exactly the offset round, even if the internal counter has already advanced.

Test Plan

Unit Tests + WOMM (both on empty / non-empty DB)

How to test locally

(using Makefile and run_runner.py of #5450)

❯ make debug-blockgen
...
❯ make run-runner
...
Running test for configuration 'test_config.yml'
generator serving on localhost:11112
Received round request 0, but already at virtualRound 1 (== g.round + g.roundOffset == 1 + 0). Not finishing round.
exiting block generator runner: signal: interrupt
Done running tests!

❯ make enter-pg
...
psql (14.5 (Debian 14.5-1.pgdg110+1))

generator_db=# select count(*) from account;
...
generator_db=# select count(*) from account_asset;
...
generator_db=# select count(*) from block_header;
...

[codecov]

Codecov Report

Merging #5449 (c2a3100) into master (1994341) will decrease coverage by 0.02%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #5449      +/-   ##
==========================================
- Coverage   55.47%   55.45%   -0.02%     
==========================================
  Files         447      447              
  Lines       63290    63290              
==========================================
- Hits        35108    35096      -12     
- Misses      25800    25820      +20     
+ Partials     2382     2374       -8     

see 16 files with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[tzaffi]

recent consensus upgrade

[winder]

I'd just like to see the one comment added. Otherwise LGTM. Thanks for the explanation during standup.

[winder]

Left some comments, the conditions here were hard for me to follow

[winder]

I'm finding this expression very difficult to understand. I think this expresses the same thing with fewer conditions

Suggested change

	if round+1 < virtualRound || virtualRound < round {
	// return cached round if available
	if round - g.roundOffset + 1 == g.round && len(g.latestBlockMsgp) != 0{
	_, err := output.Write(g.latestBlockMsgp)
	if err != nil {
	return err
	}
	}
	// "genesis" round
	if round-g.roundOffset == 0 {
	}
	// previous logic

[tzaffi]

Interestingly, for me, it's more difficult to grok this new approach. I guess each of our brains has different embedded neural nets for such inequalities.

[tzaffi]

Trying to integrate this approach landed at a compromise, hopefully clearer. Introducing nextRound and cachedRound variables: 3c4fc1c

[winder]

ty 👍