Improve performance of copy_if_else for long strings

[davidwendt]

Description

Reworks the cudf::strings::detail::copy_if_else() to improve performance for long strings. The rework builds a vector of rows to pass to the make_strings_column factory that uses the optimized gather_chars function.
Also includes a benchmark for copy_if_else specifically for strings columns.

Closes #15014

Checklist

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

[davidwendt]

Benchmark results show significant improvement for long strings.

# copy_if_else

## [0] Quadro GV100

|  row_width  |  num_rows  |   Ref Time |   Cmp Time |           Diff |   %Diff |
|-------------|------------|------------|------------|----------------|---------|
|     32      |    4096    |  81.672 us |  92.342 us |      10.670 us |  13.06% |
|     64      |    4096    |  83.276 us |  96.514 us |      13.239 us |  15.90% |
|     128     |    4096    | 108.554 us | 123.205 us |      14.650 us |  13.50% |
|     256     |    4096    | 176.912 us |  74.858 us |    -102.054 us | -57.69% |
|     512     |    4096    | 329.840 us |  76.111 us |    -253.728 us | -76.92% |
|    1024     |    4096    | 621.637 us |  80.308 us |    -541.328 us | -87.08% |
|    2048     |    4096    |   1.194 ms |  87.369 us |   -1107.081 us | -92.69% |
|    4096     |    4096    |   2.349 ms | 102.037 us |   -2246.858 us | -95.66% |
|     32      |   32768    |  78.397 us |  93.497 us |      15.100 us |  19.26% |
|     64      |   32768    | 100.368 us | 114.416 us |      14.047 us |  14.00% |
|     128     |   32768    | 149.653 us | 164.220 us |      14.567 us |   9.73% |
|     256     |   32768    | 268.554 us |  99.805 us |    -168.749 us | -62.84% |
|     512     |   32768    | 487.175 us | 109.880 us |    -377.295 us | -77.45% |
|    1024     |   32768    | 928.489 us | 134.407 us |    -794.082 us | -85.52% |
|    2048     |   32768    |   1.803 ms | 191.322 us |   -1611.378 us | -89.39% |
|    4096     |   32768    |   3.517 ms | 306.826 us |   -3210.370 us | -91.28% |
|     32      |   262144   | 147.473 us | 181.905 us |      34.432 us |  23.35% |
|     64      |   262144   | 299.250 us | 340.482 us |      41.232 us |  13.78% |
|     128     |   262144   |   1.736 ms |   1.882 ms |     145.620 us |   8.39% |
|     256     |   262144   |   8.997 ms | 315.059 us |   -8681.495 us | -96.50% |
|     512     |   262144   |  25.837 ms | 390.189 us |  -25446.713 us | -98.49% |
|    1024     |   262144   |  57.680 ms | 583.066 us |  -57097.086 us | -98.99% |
|    2048     |   262144   | 127.125 ms |   1.025 ms | -126099.652 us | -99.19% |
|    4096     |   262144   | 259.757 ms |   1.956 ms | -257800.617 us | -99.25% |
|     32      |  2097152   | 672.168 us | 841.930 us |     169.763 us |  25.26% |
|     64      |  2097152   |   1.826 ms |   2.008 ms |     182.080 us |   9.97% |
|     128     |  2097152   |  16.107 ms |  18.494 ms |       2.387 ms |  14.82% |
|     256     |  2097152   |  79.610 ms |   1.784 ms |  -77825.162 us | -97.76% |
|     512     |  2097152   | 222.464 ms |   2.418 ms | -220045.145 us | -98.91% |
|     32      |  16777216  |   4.860 ms |   5.986 ms |       1.126 ms |  23.17% |
|     64      |  16777216  |  13.876 ms |  15.054 ms |       1.178 ms |   8.49% |

Small strings suffer a bit but most are within 1-2 ms of the baseline.

[revans2]

This looks great thanks so much!

[bdice]

This is nice! It also makes the code simpler.

[vuule]

Great change: higher throughput, less code!
Some minor suggestions.

[davidwendt]

/merge