Fix an unbalanced release of the producer's pending semaphore

[erobot]

Motivation

Current code releases the producer's pending semaphore twice when batch is off and message is too big. The unbalanced release overflows the semaphore, and subsequent sends will fail with ProducerQueueIsFull.

Modifications

Remove the redundant semaphore release as the necessary release will be done in handleFailedResult.

Verifying this change

• Make sure that the change passes the CI checks.

This change modified a existing test and can be verified as follows:

• ProducerTest.testMaxMessageSize

Documentation

• doc-required
  (Your PR needs to update docs and you will update later)

• doc-not-needed
  Bug fix.

• doc
  (Your PR contains doc changes)

• doc-complete
  (Docs have been already added)

[erobot]

Some tests failed in my forked repo but can succeed in my local environment. I don't know why but seems unrelated to this PR.

Failed CI job: https://github.com/erobot/pulsar-client-cpp/actions/runs/7740402401/job/21107366697

FAILED TESTS (8/538):
78959 ms: ./pulsar-tests AuthPluginTest.testTlsDetectPulsarSslWithInvalidBroker (try #1)
78822 ms: ./pulsar-tests AuthPluginTest.testTlsDetectHttpsWithInvalidBroker (try #1)
77618 ms: ./pulsar-tests AuthPluginTest.testTlsDetectPulsarSslWithInvalidBroker (try #2)
77661 ms: ./pulsar-tests AuthPluginTest.testTlsDetectHttpsWithInvalidBroker (try #2)
76701 ms: ./pulsar-tests AuthPluginTest.testTlsDetectPulsarSslWithInvalidBroker (try #3)
76744 ms: ./pulsar-tests AuthPluginTest.testTlsDetectHttpsWithInvalidBroker (try #3)
81038 ms: ./pulsar-tests AuthPluginTest.testTlsDetectPulsarSslWithInvalidBroker (try #4)
81082 ms: ./pulsar-tests AuthPluginTest.testTlsDetectHttpsWithInvalidBroker (try #4)

[BewareMyPower]

Some tests failed in my forked repo but can succeed in my local environment. I don't know why but seems unrelated to this PR.

Not sure if it's caused by a breaking change in Pulsar 3.1.2. The apachepulsar/pulsar:latest image was upgraded from 3.1.1 to 3.1.2 today. You can open an empty PR in your fork to see if any test failed.

[erobot]

Some tests failed in my forked repo but can succeed in my local environment. I don't know why but seems unrelated to this PR.

Not sure if it's caused by a breaking change in Pulsar 3.1.2. The apachepulsar/pulsar:latest image was upgraded from 3.1.1 to 3.1.2 today. You can open an empty PR in your fork to see if any test failed.

I opened an empty PR and the same tests failed.
Empty PR: erobot#13

[BewareMyPower]

I see. I will take a look at these failed tests today.

[BewareMyPower]

I've confirmed it's a breaking change from Pulsar 3.1.2. After I switched from Pulsar 3.1.2 to Pulsar 3.1.1, the tests passed.

[BewareMyPower]

Could you rebase to master so that the CI won't be affected?

[erobot]

Could you rebase to master so that the CI won't be affected?

Rebased.