Reduce allocation cost reporting spans

[geoand]

This is done by leveraging Vert.x's Buffer
directly instead of relying on a
ByteArrayOutputStream

[quarkus-bot]

/cc @brunobat (opentelemetry), @radcortez (opentelemetry)

[brunobat]

Just to better understand this change... Will the Buffer be reused inside vert.x thus reducing the allocation needs?
The Javadoc was not clear enough.

[brunobat]

btw, main is broken. Failure in dev-ui-resources is unrelated.

[geoand]

Will the Buffer be reused inside vert.x thus reducing the allocation needs?
The Javadoc was not clear enough.

Inside Vert.x / Netty, it will.

However currently, the buffer is not pooled, meaning that each reporting of Spans will result in the allocation of a new memory segment.
If we decide to forgo allocations completely in this piece of code (cc @franz1981 @cescoffier), then this code makes it trivial, as all we have to do use the proper allocator (and guard against any exception)

[geoand]

btw, main is broken. Failure is unrelated.

Yup, unfortunately

[quarkus-bot]

Failing Jobs - Building d603488

Status	Name	Step	Failures	Logs	Raw logs
✔️	JVM Tests - JDK 11
✔️	JVM Tests - JDK 17
✖	JVM Tests - JDK 17 Windows	Build	Failures	Logs	Raw logs
✔️	JVM Tests - JDK 20

Failures

⚙️ JVM Tests - JDK 17 Windows #

- Failing: extensions/vertx-http/deployment 
! Skipped: extensions/agroal/deployment extensions/amazon-lambda-http/deployment extensions/amazon-lambda-rest/deployment and 358 more

📦 extensions/vertx-http/deployment

✖ Failed to execute goal org.apache.maven.plugins:maven-surefire-plugin:3.1.2:test (default-test) on project quarkus-vertx-http-deployment: There was a timeout in the fork