docs-build: switch to RTX Pro 6000, add changed-files filter

[jameslamb]

Contributes to rapidsai/build-planning#333

Updates to hopefully reduce how often docs-build jobs get stuck waiting for CI runners.

• switches those jobs to RTX Pro 6000 runners
• uses shared docs-build matrix to set job details

Also noticed that there's a docs_build group in the changed-files configuration, but it isn't used to limit when docs builds runs. This adds that as well.

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[jameslamb]

This looked like an omission, given that there's configuration for it:

cuvs/.github/workflows/pr.yaml

Lines 86 to 89 in 29e1101

	uses: rapidsai/shared-workflows/.github/workflows/changed-files.yaml@main
	with:
	files_yaml: |
	build_docs:

This fixes that, so some types of PRs that don't affect the docs will not have a docs CI job.

[jameslamb]

I was surprised to not see any conda packages needed for the docs builds, but looks like that was an intentional change in #2067

[jameslamb]

/merge

[jameslamb]

One C++ test failure:

[ RUN      ] AnnMGTest/AnnMGTestF_float.AnnMG/8
terminate called after throwing an instance of 'raft::cuda_error'
  what():  CUDA error encountered at: file=/tmp/conda-bld-output/bld/rattler-build_libcuvs-headers/host_env_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_placehold_/include/raft/util/cudart_utils.hpp line=139 function=void raft::copy(Type*, const Type*, size_t, rmm::_RMM_26_12::cuda_stream_view) [with Type = long int; size_t = long unsigned int]: call='cudaMemcpyAsync(dst, src, len * sizeof(Type), cudaMemcpyDefault, stream.get())', Reason=cudaErrorMisalignedAddress:misaligned address
Obtained 16 stack frames
#1 in /opt/conda/envs/test/bin/gtests/libcuvs/../../../lib/libcuvs.so(+0x6707f6) [0x7dc83781f7f6]
#2 in /opt/conda/envs/test/bin/gtests/libcuvs/../../../lib/libcuvs.so(+0x1c6d413) [0x7dc838e1c413]
#3 in /opt/conda/envs/test/bin/gtests/libcuvs/../../../lib/libcuvs.so(+0x1c6d69f) [0x7dc838e1c69f]
#4 in /opt/conda/envs/test/bin/gtests/libcuvs/../../../lib/libgomp.so.1: GOMP_parallel +0x46 [0x7dc82717ef36]
#5 in /opt/conda/envs/test/bin/gtests/libcuvs/../../../lib/libcuvs.so(+0x1c755d6) [0x7dc838e245d6]
#6 in ./../../..//bin/gtests/libcuvs/NEIGHBORS_MG_TEST(+0x70796) [0x5eed6ff16796]
#7 in ./../../..//bin/gtests/libcuvs/NEIGHBORS_MG_TEST(+0x9752c) [0x5eed6ff3d52c]
#8 in ./../../..//bin/gtests/libcuvs/NEIGHBORS_MG_TEST(+0x97801) [0x5eed6ff3d801]
#9 in ./../../..//bin/gtests/libcuvs/NEIGHBORS_MG_TEST(+0x97ba7) [0x5eed6ff3dba7]
#10 in ./../../..//bin/gtests/libcuvs/NEIGHBORS_MG_TEST(+0x98026) [0x5eed6ff3e026]
#11 in ./../../..//bin/gtests/libcuvs/NEIGHBORS_MG_TEST(+0x9eac3) [0x5eed6ff44ac3]
#12 in ./../../..//bin/gtests/libcuvs/NEIGHBORS_MG_TEST(+0x9efb1) [0x5eed6ff44fb1]
#13 in ./../../..//bin/gtests/libcuvs/NEIGHBORS_MG_TEST(+0x2eddf) [0x5eed6fed4ddf]
#14 in /usr/lib/x86_64-linux-gnu/libc.so.6(+0x2a601) [0x7dc826b5e601]
#15 in /usr/lib/x86_64-linux-gnu/libc.so.6: __libc_start_main +0x88 [0x7dc826b5e718]
#16 in ./../../..//bin/gtests/libcuvs/NEIGHBORS_MG_TEST(+0x2eece) [0x5eed6fed4ece]

CMake Error at run_gpu_test.cmake:26 (execute_process):
  execute_process failed command indexes:

    1: "Abnormal exit with child return code: Subprocess aborted"
....
The following tests FAILED:
	 32 - NEIGHBORS_MG_TEST (Failed)
/__w/cuvs/cuvs

(build link)

Re-trying.

cc @aamijar for your tracking

[jameslamb]

It's been 3.5 hours and that re-run of the C++ jobs is still going 🙃

25/38 Test #25: NEIGHBORS_ALL_NEIGHBORS_TEST ...............   Passed  1250.41 sec
      Start 26: NEIGHBORS_DYNAMIC_BATCHING_TEST

(build link)

I'm admin-merging this. It's highly unlikely that this change to the docs-building job could cause any failures in C++ tests.