[FEA] inspect wheel contents, remove anything unnecessary

[jameslamb]

Is your feature request related to a problem? Please describe.

Recent libcuopt-cu13 wheels package more than 19,000 files.

And it looks like there are many things in there that I wouldn't expect to find in a wheel for a C++ shared library, like 50MB of HTML and 27MB of PNG files, .bat files, etc.

file size
  * compressed size: 0.738G
  * uncompressed size: 1.006G
  * compression space saving: 26.6%
contents
  * directories: 2465
  * files: 19408 (41 compiled)
size by extension
  * .so - 0.438G (43.5%)
  * .pack - 0.226G (22.5%)
  * .html - 50.317M (4.9%)
  * .hpp - 43.327M (4.2%)
  * .h - 39.824M (3.9%)
  * .a - 37.721M (3.7%)
  * .3 - 31.537M (3.1%)
  * .png - 27.33M (2.7%)
  * .cu - 22.121M (2.1%)
  * .cuh - 18.394M (1.8%)
  * .o - 16.885M (1.6%)
  * .idx - 9.856M (1.0%)
  * .cpp - 9.027M (0.9%)
...

(build link)

Describe the solution you'd like

Please scrutinize the contents of the wheels being produced here and try to identify some files that could be omitted.

Summaries like I shared above can be obtained in the CI logs of wheel-build-* CI jobs, or directly by pip downloading-ing wheels and running pydistcheck --inspect on them.

Some places to look:

• use of install(DIRECTORY), install(FILES), or similar in CMake code (install(TARGETS) and dependency-tracking should be preferred)
• MANIFEST.in rules
• package_date configuration in pyproject.toml / setup.py
• third-party dependencies being vendored instead of re-used from wheels (e.g., probably do not need to vendor the raft/ headers, a runtime dependency on libraft-cu{12,13} might be enough)

After eliminating as much data as possible, look at the reported compressed sizes and reduce thresholds like this:

cuopt/python/libcuopt/pyproject.toml

Lines 64 to 68 in 933d810

	[tool.pydistcheck]
	select = [
	"distro-too-large-compressed",
	]
	max_allowed_size_compressed = '900M'

Describe alternatives you've considered

N/A

Additional context

Smaller wheels, with fewer individual files, would mean:

• faster builds
• faster installation (in CI and for users)
• smaller disk footprint for environments

[github-actions]

🔔 Hi @anandhkb @tmckayus, this issue has had no activity for 21 days. Please update or let us know if it can be closed. Thank you!

If this is an "epic" issue, then please add the "epic" label to this issue.
If it is a PR and not ready for review, then please convert this to draft.
If you just want to switch off this notification, then use the "skip inactivity reminder" label.