[CI] Avoid the use of full trace with pytest. (#11318)

[doc] Move model slicing document for R. (#11316)

Store categories from iterator. (#11313)

Store categories from cuDF. (#11311)

Test dataset for dask dataframe with str columns. (#11310)

Store categories from pandas. (#11303)

Update CUDA version for clang-tidy. (#11306)

Bump the development version to 3.1. (#11304)

[backport][dask] Fix with the latest Dask. (#11291) (#11302)

[backport][pyspark] Support columnar input for cpu pipeline (#11299) (#…

[dask] Fix with the latest Dask. (#11291)

[pyspark] Support columnar input for cpu pipeline (#11299)

Implement the container for categories. (#11297)

[BP] Update release scripts (#11289) (#11296)

[MT] Add device storage to multi-target tree. (#11277)

Cleanup PyPI variant script. (#11294)

Update release scripts (#11289)

Initial implementation of the ordinal recoder. (#11098)

Bump version to 3.0.0rc1. (#11287)

Fix init with scale pos weight. (#11280)

Fix init with scale pos weight. (#11280)

Optional normalization for the ranknet loss. (#11272)

[jvm-packages] Supports external memory (#11186)

Use parallel sort for quantile calculation when appropriate. (#11275)

Optimize device regex, test reusing a ProxyDMatrix. (#11273)

Move feature weight to skl parameters. (#9506)

Correct computation of metrics in multi-quantile regression. (#11279)

Fix timer nvtx annotation for distributed training. (#11276)

[doc] Mention the integer type for some parameters. (#11266)

[doc] Mention HPO frameworks in tuning doc. (#11253)