Update DeltaArray to always pad values to 1024 elements and use bit transpose functions

[robert3005]

This pr aims to fix the long standing issue with pushing down computation over
DecimalArrays where their validity wasn't matching the value layout. Now the
validity is also transposed. At the same time we avoid scalar encoding and pad
values to 1024 elements

fix #4973

Signed-off-by: Robert Kruszewski github@robertk.io

[codspeed-hq]

Merging this PR will degrade performance by 15.51%

❌ 2 regressed benchmarks
✅ 1014 untouched benchmarks
⏩ 1522 skipped benchmarks¹

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	bitwise_not_vortex_buffer_mut[128]	317.8 ns	376.1 ns	-15.51%
❌	Simulation	bitwise_not_vortex_buffer_mut[1024]	477.2 ns	535.6 ns	-10.89%

---

_{Comparing rk/delta (ad7d841) with develop (43ae2dc)}

Footnotes

1. 1522 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[robert3005]

We should add unstable_encodings to the fuzzer

[joseph-isaacs]

Nice just a q about the validity padding

[joseph-isaacs]

Good to know that (CI / Compat check / compat-test (pull_request)) works, but we can replace that array going forwards

[robert3005]

I have made the transpose logic use the buffer in place where possible and asserted that untranspose must have been already padded to multiple of 128

[robert3005]

I have gotten rid of delta fixture and the backcompat checking. We can revive this once we are happy with the format of it.

[joseph-isaacs]

very nice