Fix flaky flexbuffers_alloc_check test in cargo test

[RenzoMXD]

The alloc check binaries use a global non-thread-safe allocation counter (static mut NUM_ALLOCS) to assert zero heap allocations on warm builder reuse.
When run via cargo test, the test harness's background threads occasionally allocate, inflating the counter and causing intermittent assertion failures. These binaries are already run reliably via cargo run --bin in RustTest.sh, so disabling them as test targets avoids the flaky behavior with no loss of coverage.
Closes #8961

[jtdavis777]

Just to make sure I understand your comment, these tests are executed as part of a PR build elsewhere? What is the use of this file?

Thank you for all your help :)

[RenzoMXD]

Just to make sure I understand your comment, these tests are executed as part of a PR build elsewhere? What is the use of this file?

RustTest.sh runs two commands that both execute flexbuffers_alloc_check:

1. Line 46: cargo test -- --quiet — This runs all test targets in the package, including any [[bin]] with #[test] functions. It picks up flexbuffers_alloc_check because it has a #[test] fn no_extra_allocations().

2. Line 55: cargo run --bin=flexbuffers_alloc_check — This runs the same binary directly via its main() function, without the test harness.

So the same binary was being executed twice in a single CI run — first by cargo test (with the test harness, which adds background threads and extra allocations), then by cargo run --bin (clean, no harness). The flaky failure happens in step 1 because the test harness threads' allocations interfere with the global counter.

With test = false, cargo test skips these binaries entirely, and they only run via the reliable cargo run --bin path.