Simplify OrderPreservingInterner allocation strategy ~97% faster (#2677)

[tustvold]

Which issue does this PR close?

Closes #2677

Rationale for this change

Not only was the previous allocation strategy quite hard to follow, when interning sorted data it would end up comparing the value against at least 1/4 of the already interned values. This was sub-optimal.

Taking a step back the original approach of leaving spaces was to mitigate the impact of unsorted data, as the design evolved OrderPreservingInterner now always sorts the dictionary prior to interning. This dramatically reduces the likelihood of out-of-order inserts, and consequently diminishes the value of this mitigation.

A simpler, in-order allocation strategy is both faster and easier to understand

Benchmarking convert_columns 4096 string_dictionary(10, 0): Warming up for 3.0000 s
Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 5.6s, enable flat sampling, or reduce sample count to 60.
convert_columns 4096 string_dictionary(10, 0)
                        time:   [1.1147 ms 1.1156 ms 1.1169 ms]
                        change: [-96.615% -96.612% -96.607%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 100 measurements (3.00%)
  1 (1.00%) high mild
  2 (2.00%) high severe

Benchmarking convert_columns 4096 string_dictionary(30, 0): Warming up for 3.0000 s
Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 5.8s, enable flat sampling, or reduce sample count to 60.
convert_columns 4096 string_dictionary(30, 0)
                        time:   [1.1510 ms 1.1514 ms 1.1519 ms]
                        change: [-96.623% -96.621% -96.619%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 100 measurements (2.00%)
  2 (2.00%) high mild

Benchmarking convert_columns 4096 string_dictionary(100, 0): Warming up for 3.0000 s
Warning: Unable to complete 100 samples in 5.0s. You may wish to increase target time to 6.1s, enable flat sampling, or reduce sample count to 60.
convert_columns 4096 string_dictionary(100, 0)
                        time:   [690.29 µs 692.36 µs 695.23 µs]
                        change: [-97.493% -97.335% -97.211%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 20 outliers among 100 measurements (20.00%)
  1 (1.00%) low mild
  19 (19.00%) high severe

convert_columns 4096 string_dictionary(100, 0.5)
                        time:   [316.77 µs 316.89 µs 317.00 µs]
                        change: [-95.710% -95.708% -95.705%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 100 measurements (3.00%)
  1 (1.00%) high mild
  2 (2.00%) high severe

What changes are included in this PR?

Changes the allocation strategy for OrderPreservingInterner to be simpler to understand and faster.

Are there any user-facing changes?

No

[alamb]

I read the code carefully and it makes sense to me. I had one question about test coverage but otherwise 👌

I look forward to seeing how this works

[alamb]

Is there coverage of this branch? I think it happens when data is inserted in backwards sorted order, right? Maybe we can add at test to ensure it is covered?

[tustvold]

This branch is hit when the current bucket is "full" and it is trying to insert a value that is greater than the last value in the bucket. This therefore occurs when interning more than 255 values in order, and is therefore hit by test_interner() in particular

let mut values: Vec<_> = (0_u64..2000).collect();
test_intern_values(&values);

I have confirmed this with a debugger

[ursabot]

Benchmark runs are scheduled for baseline = f4ee8b9 and contender = 65d5576. 65d5576 is a master commit associated with this PR. Results will be available as each benchmark for each run completes.
Conbench compare runs links:
[Skipped ⚠️ Benchmarking of arrow-rs-commits is not supported on ec2-t3-xlarge-us-east-2] ec2-t3-xlarge-us-east-2
[Skipped ⚠️ Benchmarking of arrow-rs-commits is not supported on test-mac-arm] test-mac-arm
[Skipped ⚠️ Benchmarking of arrow-rs-commits is not supported on ursa-i9-9960x] ursa-i9-9960x
[Skipped ⚠️ Benchmarking of arrow-rs-commits is not supported on ursa-thinkcentre-m75q] ursa-thinkcentre-m75q
Buildkite builds:
Supported benchmarks:
ec2-t3-xlarge-us-east-2: Supported benchmark langs: Python, R. Runs only benchmarks with cloud = True
test-mac-arm: Supported benchmark langs: C++, Python, R
ursa-i9-9960x: Supported benchmark langs: Python, R, JavaScript
ursa-thinkcentre-m75q: Supported benchmark langs: C++, Java