Split up mixed-join kernels source files

[davidwendt]

Split up mixed_join_kernels.cu and mixed_join_size_kernels.cu to improve overall build time.
Currently these take about 30 minutes each on the gpuCI build. Example of a recent build metrics report:
https://gpuci.gpuopenanalytics.com/job/rapidsai/job/gpuci-22-06/job/cudf/job/prb/job/cudf-cpu-cuda-build/CUDA=11.5/164/Build_20Metrics_20Report/

The nulls and non-nulls definitions are placed into separate source files.

The kernel source used for mixed_join_kernels.cu (both null and non-null) is moved to mixed_join_kernel.cuh and the nulls definition is moved to mixed_join_kernel_nulls.cu. For consistency the mixed_join_kernels.cu name is changed to just mixed_join_kernel.cu since it now only contains one definition.
This same pattern applies to mixed_join_size_kernels.cu splitting into mixed_join_size_kernel.cuh, mixed_join_size_kernel_nulls.cu and mixed_join_size_kernel.cu

No function behavior or actual code generation has changed. The source code has just moved into more source files to help better parallelize and speed up the build process. This improves compile time by 10% for a release build and ~25% for a debug build.

[codecov]

Codecov Report

Merging #10671 (587cb1b) into branch-22.06 (94a5d41) will increase coverage by 0.02%.
The diff coverage is n/a.

@@               Coverage Diff                @@
##           branch-22.06   #10671      +/-   ##
================================================
+ Coverage         86.38%   86.41%   +0.02%     
================================================
  Files               142      142              
  Lines             22334    22334              
================================================
+ Hits              19294    19300       +6     
+ Misses             3040     3034       -6     

Impacted Files	Coverage Δ
python/cudf/cudf/core/column/string.py	89.22% <0.00%> (+0.12%)	⬆️
python/cudf/cudf/core/groupby/groupby.py	91.72% <0.00%> (+0.22%)	⬆️
python/cudf/cudf/core/tools/datetimes.py	84.49% <0.00%> (+0.30%)	⬆️
python/cudf/cudf/core/column/lists.py	92.79% <0.00%> (+1.27%)	⬆️

---

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 9409559...587cb1b. Read the comment docs.

[davidwendt]

Here are the times for mixed_join_kernels.cu and mixed_join_size_kernels.cu with but nulls and non-nulls templated functions defined in single source files:
https://gpuci.gpuopenanalytics.com/job/rapidsai/job/gpuci-22-06/job/cudf/job/prb/job/cudf-cpu-cuda-build/CUDA=11.5/323/Build_20Metrics_20Report/

Here are the times after splitting up each of these into nulls and non-nulls source files:
https://gpuci.gpuopenanalytics.com/job/rapidsai/job/gpuci-22-06/job/cudf/job/prb/job/cudf-cpu-cuda-build/CUDA=11.5/329/Build_20Metrics_20Report/

This improved the overall build time by about 10 minutes. And though this may be only a 10% improvement for a release build, this change reduces a debug build from 2 hours to 1.5 hours (~25%).

[robertmaynard]

CMake changes LGTM

[bdice]

Thanks @davidwendt!

[davidwendt]

@gpucibot merge