[1.x] Backport Add cmake flag USE_FATBIN_COMPRESSION, ON by default (#19123)

[DickJC123]

Description

Backport of PR #19123. Reduces the size of nvcc fatbin output files via compression. Helps libmxnet.so to stay within 2GB limit that causes linker failures.

Checklist

Essentials

• PR's title starts with a category (e.g. [BUGFIX], [MODEL], [TUTORIAL], [FEATURE], [DOC], etc)
• Changes are complete (i.e. I finished coding on this PR)
• All changes have test coverage
• Code is well-documented

Changes

• Feature1, tests, (and when applicable, API doc)
• Feature2, tests, (and when applicable, API doc)

Comments

• If this change is a backward incompatible change, why must this change be made.
• Interesting edge cases to note here

[mxnet-bot]

Hey @DickJC123 , Thanks for submitting the PR
All tests are already queued to run once. If tests fail, you can trigger one or more tests again with the following commands:

• To trigger all jobs: @mxnet-bot run ci [all]
• To trigger specific jobs: @mxnet-bot run ci [job1, job2]

---

CI supported jobs: [windows-cpu, website, windows-gpu, clang, sanity, centos-cpu, miscellaneous, edge, centos-gpu, unix-gpu, unix-cpu]

---

Note:
Only following 3 categories can trigger CI :PR Author, MXNet Committer, Jenkins Admin.
All CI tests must pass before the PR can be merged.

[mxnet-bot]

Unauthorized access detected.
Only following 3 categories can trigger CI :
PR Author, MXNet Committer, Jenkins Admin.

[samskalicky]

It looks like CMAKE_CUDA_FLAGS gets set again later in the windows flow:
https://github.com/apache/incubator-mxnet/blob/4bb3df00ba5c509975363da8ece5cd89dc8f574a/CMakeLists.txt#L725
wont this overwrite the previous fatbin setting?

[DickJC123]

I've pushed a fix to this set-vs-string_append, and changed a few minor details with the approach. I'd be surprised if it fixes things since the referenced line is not in the build's code path (I think) due to:

https://github.com/apache/incubator-mxnet/blob/38c9d57f9a14578be4fc739b9d6880b49b00c4fc/CMakeLists.txt#L76

Is it possible to have the 1.x Windows build use the same MSVC compiler as master (which doesn't seem to have the problem with -compress-all or the link OOM error)?