[redis_proxy] Assert proper transaction initialization

[dbarbosapn]

Fixes #37825

Making sure we know early that transaction initialization was not done correctly.

Tested manually, was unable to reproduce issue again. I'm guessing the assertion caused compiler to not do bad optimization there 🤷

[mattklein123]

Sorry this doesn't make sense. We need to better understand the problem. Are you sure this path is properly covered in test and by ASAN/TSAN?

Also use the Envoy ASSERT macro which is not compiled on release builds, which if this is an optimization issue (very doubtful) won't help.

/wait

[dbarbosapn]

Sorry this doesn't make sense. We need to better understand the problem. Are you sure this path is properly covered in test and by ASAN/TSAN?

Also use the Envoy ASSERT macro which is not compiled on release builds, which if this is an optimization issue (very doubtful) won't help.

/wait

Oh yeah it doesn't make any sense 😂 Just created PR to have the CI build to see if it wasn't just local the issue (if the assertion would fire).

I'll keep investigating. Maybe it was something transient on that specific build.

Will try tomorrow latest main, to see if the issue is there or not. If not, I'll abandon this PR.

[dbarbosapn]

Closing as not able to reproduce with latest binaries. Will reopen investigation if we notice this again 😃