Fix out-of-memory error in compiled-binaryop benchmark

[davidwendt]

Fixes out-of-memory error that occurs when running the BINARYOP_BENCH COMPILED_BINARYOP benchmark combined with the BINARYOP benchmark. The following minimal command shows the error:

benchmarks/BINARYOP_BENCH '--benchmark_filter=COMPILED_BINARYOP|100000000/10'
...
BINARYOP<double, TreeType::IMBALANCED_LEFT, false>/binaryop_double_imbalanced_unique/100000000/10/manual_time                              40.4 ms         40.4 ms           17 bytes_per_second=202.953G/s
terminate called after throwing an instance of 'rmm::out_of_memory'
  what():  std::bad_alloc: out_of_memory: RMM failure at:/conda/envs/rapids/include/rmm/mr/device/pool_memory_resource.hpp:192: Maximum pool size exceeded
Aborted (core dumped)

The COMPILED_BINARYOP is using a TEMPLATED_BENCHMARK_F macro which causes a new separate memory pool to be created instead of reusing the one already created by BINARYOP.
This PR reworks the benchmark macros in compiled_binaryop.cpp to avoid using TEMPLATED_BENCHMARK_F allowing it share the existing memory pool.

Similar to #10258

[codecov]

Codecov Report

Merging #10269 (bbb3282) into branch-22.04 (eb5e3e3) will not change coverage.
The diff coverage is n/a.

@@              Coverage Diff              @@
##           branch-22.04   #10269   +/-   ##
=============================================
  Coverage         10.43%   10.43%           
=============================================
  Files               122      122           
  Lines             20583    20583           
=============================================
  Hits               2147     2147           
  Misses            18436    18436           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update eb5e3e3...bbb3282. Read the comment docs.

[mythrocks]

Thank you. TIL.

[karthikeyann]

I too fixed the same bug locally with same fix.
Thanks for doing this.

TEMPLATED_BENCHMARK_F macro which causes a new separate memory pool

Still unsure how this is caused.

[karthikeyann]

LGTM 👍

[karthikeyann]

@gpucibot merge