Partial cuIO GPU decompression refactor

[vuule]

Required to expand future nvcomp integration.

• Moving nvcomp integration in ORC and Parquet readers to common code. Enables nvcomp use for multiple compression type without code duplication.
• gpu_inflate_input_s refactor to facilitate unified host/device decompressor interface. Enables further changes to unify CPU and GPU decompression API, which in turn enables ZSTD use in ORC.

[codecov]

Codecov Report

Merging #10699 (f0f389e) into branch-22.06 (dc1435b) will increase coverage by 0.06%.
The diff coverage is 92.42%.

@@               Coverage Diff                @@
##           branch-22.06   #10699      +/-   ##
================================================
+ Coverage         86.37%   86.43%   +0.06%     
================================================
  Files               142      143       +1     
  Lines             22306    22444     +138     
================================================
+ Hits              19266    19399     +133     
- Misses             3040     3045       +5     

Impacted Files	Coverage Δ
python/cudf/cudf/_lib/__init__.py	100.00% <ø> (ø)
python/dask_cudf/dask_cudf/io/parquet.py	92.39% <81.08%> (-1.40%)	⬇️
python/cudf/cudf/core/_internals/expressions.py	92.85% <92.85%> (ø)
python/cudf/cudf/core/dataframe.py	93.74% <93.33%> (-0.01%)	⬇️
python/cudf/cudf/core/column/lists.py	92.91% <100.00%> (+1.25%)	⬆️
python/cudf/cudf/core/groupby/groupby.py	91.79% <100.00%> (+0.34%)	⬆️
python/cudf/cudf/core/indexed_frame.py	91.70% <100.00%> (ø)
...ython/dask_cudf/dask_cudf/io/tests/test_parquet.py	100.00% <100.00%> (ø)
python/cudf/cudf/core/column/string.py	89.21% <0.00%> (+0.12%)	⬆️
... and 2 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 9ac2477...f0f389e. Read the comment docs.

[bdice]

The improvements from reviews made this PR look a lot better! I have some comments attached and will approve after some of the comments are addressed.

An apology in advance -- I attempted to find/fix some places where docstrings seemed to be marking inputs as @param[out]. While this may in fact be correct in some of the functions, I'm not sure that all the docstrings are correct -- and I'm also not sure if I correctly fixed all of the errors. Similarly, there are a few device_span<device_span<T const>> that could probably be device_span<device_span<T const> const> (with an extra const on the outer span). I couldn't quite track all the functions' input/output patterns from what I saw in the diffs, and I decided it was not worth digging deeper to verify each one at the cost of delaying the review. I would suggest a very quick pass to double-check your expectations for constness of inputs/outputs/statuses.

[elstehle]

Almost through. Just a few minor comments and some nitpicks.

[elstehle]

Thanks for your effort, @vuule! LGTM, once remaining comments are addressed.

[bdice]

Discussed offline a bit with @vuule. The changes in 2bd4f4e are nice. 👍

[vuule]

@gpucibot merge