Add CUDA decompression kernels for OnPair

[robert3005]

They follow the same pattern as FSST where we let users choose varbin/varbinview

[codspeed-hq]

Merging this PR will improve performance by 30.31%

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 3 improved benchmarks
✅ 1980 untouched benchmarks
⏩ 46 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	decode_varbin[(1000, 4)]	99.3 µs	60.1 µs	+65.1%
⚡	Simulation	cold_misaligned[(64, 256)]	5.3 ms	4.4 ms	+20.73%
⚡	Simulation	compress_fsst[(1000, 64, 8)]	1.2 ms	1 ms	+11%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing rk/onpairgpu (47ad0ab) with develop (b4b8f54)}

Footnotes

1. 46 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]

I made it use the same machinery like #8921 to avoid triple kernel dispatch

[0ax1]

lgtm. but worth double checking all index types

[joseph-isaacs]

nno no

[robert3005]

I am reworking this to make sure all the work happens in one place - there's too many conflation. Will make sure that all non trivial compute is on gpu

[robert3005]

@joseph-isaacs have another look

[robert3005]

I have added a feature to let you spawn work on another stream here #9234. We can then rebase this