Increase DeltaBitPackEncoder miniblock size to 64 for 64-bit integers (#2282)

[tustvold]

Which issue does this PR close?

Part of #2282

Rationale for this change

Following #2278 we can efficiently decode blocks of 64x 64-bit integers, however, the defaults of DeltaBitPackEncoder configure 32-wide miniblocks. #2282 tracks allowing users to configure this, but in the short term we should choose a more optimal default.

arrow_array_reader/Int64Array/binary packed, mandatory, no NULLs                                                                             
                        time:   [25.295 us 25.307 us 25.322 us]
                        change: [-59.630% -59.504% -59.394%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 8 outliers among 100 measurements (8.00%)
  5 (5.00%) high mild
  3 (3.00%) high severe
arrow_array_reader/Int64Array/binary packed, optional, no NULLs                                                                             
                        time:   [41.864 us 41.871 us 41.878 us]
                        change: [-47.152% -47.076% -46.954%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 9 outliers among 100 measurements (9.00%)
  2 (2.00%) low mild
  3 (3.00%) high mild
  4 (4.00%) high severe
arrow_array_reader/Int64Array/binary packed, optional, half NULLs                                                                             
                        time:   [46.301 us 46.324 us 46.345 us]
                        change: [-26.069% -25.818% -25.590%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 100 measurements (2.00%)
  1 (1.00%) high mild

What changes are included in this PR?

Changes to using a miniblock size of 64 for 64-bit integers

Are there any user-facing changes?

There should be no breaking user-facing changes, although this may change the way their data is encoded.

[tustvold]

For comparison the PLAIN encoding is

arrow_array_reader/Int64Array/plain encoded, mandatory, no NULLs                                                                             
                        time:   [8.0204 us 8.1321 us 8.2710 us]

So within a factor of 3, which is cool. It's a shame this encoding isn't one of the more vectorisable delta encodings (despite the docs linking to a paper that explicitly says why they are superior to the one parquet in the end chose), but it is what it is 😅

[alamb]

Can we summarize this PR as "increases performance of reading integers from parquet by 20%-60%"?

[alamb]

looks good to me, thank you @tustvold

Even though it is a small PR, I think it would be good if someone else with some more expertise in this area like @sunchao or @nevi-me could also give it a look prior to merging

[tustvold]

increases performance of reading integers from parquet by 20%-60%

It only increases the performance for 64-bit integers, encoded with DeltaBitPacking, which in turn is only used when the writer version is 2 and the dictionary encoding is either disabled or has spilled. So whilst technically true, that might be a somewhat misleading way of phrasing it 😅

[tustvold]

I'm going to get this in so that it can make the arrow 20 release

[ursabot]

Benchmark runs are scheduled for baseline = 0af81e8 and contender = 297a8fa. 297a8fa 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