[C++] make asan run faster

[chaokunyang]

Feature Request

https://github.com/apache/fory/actions/runs/23642273202/job/68865633774?pr=3522

Is your feature request related to a problem? Please describe

No response

Describe the solution you'd like

Not all tests need run, we can just run struct_serializer_test.h and smart_ptr_test and collection_test

Describe alternatives you've considered

No response

Additional context

No response

[Geethapranay1]

I investigated the asan slowness and it’s dominated by //cpp/fory/util:float16_test (~246s),
specifically Float16FromFloatTest.SignSymmetryForNonNaNBitPatterns (large float32 bit-pattern scan).

I implemented a small change that keeps normal-build coverage, but skips only this one test under ASAN
(detect ASAN via __has_feature(address_sanitizer) / __SANITIZE_ADDRESS__ and call GTEST_SKIP()).

Local verification:

• bazel test --config=asan --action_env=CCACHE_DISABLE=1 //cpp/fory/util:float16_test --test_output=all
SignSymmetryForNonNaNBitPatterns shows as [SKIPPED] under asan and warm-cache elapsed time is ~0.39s.

i have a question do you prefer this targeted approach (skip only the pathological float16 test under asan),
or would you rather change the asan workflow to run only
//cpp/fory/serialization:struct_test, :smart_ptr_serializer_test, :collection_serializer_test
(as mentioned in the issue)?
I can proceed with either (or do the targeted change first and a follow-up workflow change if needed).

[chaokunyang]

@Geethapranay1 Would you like to optimize //cpp/fory/util:float16_test to make it faster?

[Geethapranay1]

Yeah I can do that! The bottleneck is SignSymmetryForNonNaNBitPatterns it brute-forces ~41k float32 bit patterns which takes ~246s under ASAN. I'll rework it to sample the interesting regions (underflow/overflow boundaries, subnormals) instead of scanning the whole range. Will push a PR soon.