Optimise primitive clazz checks

[franz1981]

I've run few microbenchmarks to be sure it is actually better performing than the original code, which has proved (in #37656) to be too costy.

I've used franz1981/java-puzzles@aa09c3d
with a fair distribution of class types, to save the JIT to optimize it somehow.

The results with # VM version: JDK 21, Java HotSpot(TM) 64-Bit Server VM, 21+35-LTS-2513 are

Benchmark                                             (noPrimitives)   Mode  Cnt           Score          Error  Units
CheckPrimitiveWrappers.lookupPrimitiveWrapper                   true  thrpt   30  1468019504.269 ± 14740624.010  ops/s
CheckPrimitiveWrappers.lookupPrimitiveWrapper                  false  thrpt   30   916703637.882 ±  7936655.776  ops/s
CheckPrimitiveWrappers.lookupPrimitiveWrapperWithMap            true  thrpt   30   729446870.700 ±  1902697.898  ops/s
CheckPrimitiveWrappers.lookupPrimitiveWrapperWithMap           false  thrpt   30   507266225.423 ±  4544529.352  ops/s

• Closes Serialisers.findWriters primitive wrapper lookup performance #37656

[geoand]

Let's see what the perf test says

[franz1981]

Given the existing baseline is just a very tiny improvement, like 4 M req/sec -> 4.1 M req/sec but consistent, meaning that is not noise.

[geoand]

Sounds good!

[quarkus-bot]

✔️ The latest workflow run for the pull request has completed successfully.

It should be safe to merge provided you have a look at the other checks in the summary.