fix: #341 align attempt retry count with await and recover

[edmoffo]

Issue #341 noticed that BazaRb#attempt passed @retries to with_retries(max_tries:) while #await and #recover treated @retries as a retry count, so a retries: 5 configuration produced four TimedOut retries against five 429 and >=499 retries. The YARD on lib/baza-rb.rb:58 names the kwarg "Number of retries on connection failure", which matches the loop semantics in await and recover and contradicts the with_retries semantics in attempt.

The diff bumps max_tries: in attempt to @retries + 1 so all three retry layers interpret the constructor's retries: kwarg as the same retry count. The visible effect is that a BazaRb.new(host, port, token, retries: N) now performs N + 1 total tries on TimedOut, 429, and >=499 — one initial call and N retries on each path. The new test_attempt_total_tries_match_retries_plus_one in test/test_baza_edge.rb pins the count at @retries + 1 total tries on a persistent TimedOut.

Local checks on Ruby 3.3.6: bundle exec ruby -Ilib -Itest test/test_baza_edge.rb -n test_attempt_total_tries_match_retries_plus_one -> 1 test, 2 assertions, 0 failures; bundle exec ruby -Ilib -Itest test/test_baza_edge.rb -n '/retries|retry|attempt|partial_file/' -> 8 tests, 21 assertions, 0 failures; bundle exec ruby -Ilib -Itest test/test_baza_edge.rb -> 39 tests, 1 failure and 7 errors all reproducing on origin/master (test_durable_load_in_chunks, test_real_http, the four test_push_with_*/test_push_compress* cases, test_with_very_short_timeout, test_durable_load_from_sinatra) and unrelated to this diff; bundle exec rubocop lib/baza-rb.rb test/test_baza_edge.rb -> no offenses; bundle exec yard --fail-on-warning 'lib/**/*.rb' -> 96.23% documented, no warnings.

Closes #341.

[GHX5T-SOL]

Approved. This current head fixes the retry-count mismatch from #341 by treating @retries as a retry count consistently: attempt now passes max_tries: @retries + 1, matching the existing await/recover loop semantics and the constructor YARD. The added test_attempt_total_tries_match_retries_plus_one pins the timeout path at one initial try plus @retries retries, so the one-line source change is covered.

Validation I ran on exact head ddfc2c9e098a75081ade1ce534e5fe04184cd291:

• bundle check -> dependencies satisfied
• ruby -c lib/baza-rb.rb -> Syntax OK
• git diff --check upstream/master...HEAD -> passed
• focused test_attempt_total_tries_match_retries_plus_one with coverage disabled -> 1 test, 2 assertions, 0 failures
• retry/attempt subset with coverage disabled (/retries|retry|attempt|partial_file/) -> 8 tests, 21 assertions, 0 failures
• full test/test_baza_edge.rb with coverage disabled -> 39 tests, 110 assertions, 0 failures
• full test/test_fake.rb with coverage disabled -> 36 tests, 46 assertions, 0 failures
• bundle exec rubocop -> 12 files inspected, no offenses
• bundle exec rake picks -> passed
• bundle exec rake yard -> 96.23% documented

Hosted readback: the touched regression passes on both hosted rake logs. Ubuntu reached 109 tests, 200 assertions, 0 failures, 0 errors before the workflow was canceled after style/docs completion. macOS has one failure, but it is the known live TestBazaLive#test_live_full_cycle stuck-job case (Job #74847 ... did not finish) while the retry-count regression and other relevant local tests pass. I did not run local full bundle exec rake because it invokes live Zerocracy API tests, which I kept out of scope.

[0crat]

@GHX5T-SOL Hey! Nice work on the review! 🎉 You scored +6 points this time: started with the standard +18 base, but lost -8 for zero comments (our policy really values reviewer feedback!) and -4 for having only 19 hits-of-code (we need at least 24). Your running total is now +1372 - keep it up and don't forget to check your Zerocracy account for updates!