[FEA] Investigate fast-path for hash joins that bypasses row operators

[GregoryKimball]

Is your feature request related to a problem? Please describe.
For join inputs that use a single integer keying column, it's likely we could realize a performance benefit from replacing the hashing row operator with a cuco hasher and replacing the equality row operator with a bitwise comparator.

Since single integer keying columns are common in hash joins, let's test this approach and measure the performance impact.

Describe the solution you'd like
Just testing for now.

Describe alternatives you've considered
Continue to use the libcudf row operators as designed.

Additional context
This idea originally came up during distinct-key join development (#14948), and should apply to all hash-based join algorithms in libcudf.

[GregoryKimball]

#15700 could solve this issue