Improve contains_column by invoking contains_table

[PointKernel]

Description

Part of ##12261

This PR simplifies the contains_column implementation by invoking contains_table and gets rid of the use of the cudf unordered_multiset. It also removes the unordered_multiset header file from libcudf.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[PointKernel]

Performance comparison between before and after this PR shows that the new implementation can bring up to 2.2x speedups over the old for large data and can be 15% slower than the current code in the worst scenario:

Benchmark                                                             Time             CPU      Time Old      Time New       CPU Old       CPU New
--------------------------------------------------------------------------------------------------------------------------------------------------
Search/ColumnContains_AllValid/1024/manual_time                    -0.0201         -0.0147             0             0             0             0
Search/ColumnContains_AllValid/4096/manual_time                    +0.1479         +0.1027             0             0             0             0
Search/ColumnContains_AllValid/32768/manual_time                   +0.3774         +0.2879             0             0             0             0
Search/ColumnContains_AllValid/262144/manual_time                  +0.1825         +0.1631             0             0             0             0
Search/ColumnContains_AllValid/2097152/manual_time                 -0.4995         -0.4977             4             2             4             2
Search/ColumnContains_AllValid/16777216/manual_time                -0.4629         -0.4627            42            23            42            23
Search/ColumnContains_AllValid/67108864/manual_time                -0.5701         -0.5700            82            35            82            35
Search/ColumnContains_Nulls/1024/manual_time                       +0.0138         +0.0081             0             0             0             0
Search/ColumnContains_Nulls/4096/manual_time                       +0.0807         +0.0570             0             0             0             0
Search/ColumnContains_Nulls/32768/manual_time                      +0.4188         +0.3202             0             0             0             0
Search/ColumnContains_Nulls/262144/manual_time                     +0.4772         +0.4089             0             0             0             0
Search/ColumnContains_Nulls/2097152/manual_time                    -0.4655         -0.4635             4             2             4             2
Search/ColumnContains_Nulls/16777216/manual_time                   -0.4141         -0.4139            38            22            38            22
Search/ColumnContains_Nulls/67108864/manual_time                   -0.5455         -0.5453            77            35            77            35
OVERALL_GEOMEAN                                                    -0.1737         -0.1887             0             0             0             0

[bdice]

I've wanted this change for a long time! Thanks @PointKernel. The benchmarks look good for large sizes, despite the minor slowdown for medium sizes.

[PointKernel]

/merge