[nvtt] Control CUDA dependency, fix linkage

[dg0yt]

• What does your PR fix?

  Fixes a warning about the unused debug postfix in the release build.
  Fixes wrong CRT linkage for MSVC release builds.
  Adds an opt-in feature which controls the CUDA dependency.
  Fixes an installation order problem detected in a vcpkg CI build of port osg when port cuda wasn't installed, but needed by the cached nvtt package.
  Other changes to port nvtt are out of scope for this PR.

• Which triplets are supported/not supported? Have you updated the CI baseline?

  unchanged, no.

• Does your PR follow the maintainer guide?

  yes

• If you have added/updated a port: Have you run ./vcpkg x-add-version --all and committed the result?

  yes

If you are still working on the PR, open it as a Draft: https://github.blog/2019-02-14-introducing-draft-pull-requests/

[dg0yt]

Note on CI results: The binary artifact for port nvtt was already cached from build of #24720.

[dg0yt]

Now examining another osg (port update) build issue in release build only, x64-windows:

nvtt.lib(CompressionOptions.cpp.obj) : error LNK2038: mismatch detected for 'RuntimeLibrary': value 'MT_StaticRelease' doesn't match value 'MD_DynamicRelease' in NVTTImageProcessor.cpp.obj

and similar.

Looking at the logs from the forced nvtt build failure.
x64-windows debug uses /MDd:

[56/106] C:\PROGRA~1\MICROS~1\2022\ENTERP~1\VC\Tools\MSVC\1432~1.313\bin\Hostx64\x64\cl.exe   /TP -DNVTT_EXPORTS -D__MMX__ -D__SSE2__ -D__SSE__ -ID:\buildtrees\nvtt\src\2188752e3f-17c078710b.clean\src -ID:\buildtrees\nvtt\x64-windows-dbg\src -ID:\buildtrees\nvtt\src\2188752e3f-17c078710b.clean\extern\poshlib -ID:\buildtrees\nvtt\src\2188752e3f-17c078710b.clean\extern\stb -ID:\buildtrees\nvtt\src\2188752e3f-17c078710b.clean\src\nvtt -ID:\buildtrees\nvtt\src\2188752e3f-17c078710b.clean\extern\rg_etc1_v104 /nologo /DWIN32 /D_WINDOWS /W3 /utf-8 /GR /EHsc /MP  /D_DEBUG /MDd /Z7 /Ob0 /Od /RTC1  -std:c++14 /showIncludes /Fosrc\nvtt\CMakeFiles\nvtt.dir\CompressionOptions.cpp.obj /Fdsrc\nvtt\CMakeFiles\nvtt.dir\nvtt.pdb /FS -c D:\buildtrees\nvtt\src\2188752e3f-17c078710b.clean\src\nvtt\CompressionOptions.cpp

x64-windows release uses no flag but probably should use /MD?

[60/106] C:\PROGRA~1\MICROS~1\2022\ENTERP~1\VC\Tools\MSVC\1432~1.313\bin\Hostx64\x64\cl.exe   /TP -DNVTT_EXPORTS -D__MMX__ -D__SSE2__ -D__SSE__ -ID:\buildtrees\nvtt\src\2188752e3f-17c078710b.clean\src -ID:\buildtrees\nvtt\x64-windows-rel\src -ID:\buildtrees\nvtt\src\2188752e3f-17c078710b.clean\extern\poshlib -ID:\buildtrees\nvtt\src\2188752e3f-17c078710b.clean\extern\stb -ID:\buildtrees\nvtt\src\2188752e3f-17c078710b.clean\src\nvtt -ID:\buildtrees\nvtt\src\2188752e3f-17c078710b.clean\extern\rg_etc1_v104 /nologo /DWIN32 /D_WINDOWS /W3 /utf-8 /GR /EHsc /MP   /nologo /DWIN32 /D_WINDOWS /W3 /utf-8 /GR /EHsc /MP  /O2 /Ob2 /Oi /Ot /Oy /GL -std:c++14 /showIncludes /Fosrc\nvtt\CMakeFiles\nvtt.dir\CompressionOptions.cpp.obj /Fdsrc\nvtt\CMakeFiles\nvtt.dir\nvtt.pdb /FS -c D:\buildtrees\nvtt\src\2188752e3f-17c078710b.clean\src\nvtt\CompressionOptions.cpp

I see, it doesn't save the existing CMAKE_<lang>_FLAGS_RELEASE in cmake/OptimalOptions.cmake.
(I wonder why it only fails in my port osg update.)

[dg0yt]

Now there is /MD for release:

[59/106] C:\PROGRA~1\MICROS~1\2022\ENTERP~1\VC\Tools\MSVC\1432~1.313\bin\Hostx64\x64\cl.exe   /TP -DNVTT_EXPORTS -D__MMX__ -D__SSE2__ -D__SSE__ -ID:\buildtrees\nvtt\src\2188752e3f-686fd8e62c.clean\src -ID:\buildtrees\nvtt\x64-windows-rel\src -ID:\buildtrees\nvtt\src\2188752e3f-686fd8e62c.clean\extern\poshlib -ID:\buildtrees\nvtt\src\2188752e3f-686fd8e62c.clean\extern\stb -ID:\buildtrees\nvtt\src\2188752e3f-686fd8e62c.clean\src\nvtt -ID:\buildtrees\nvtt\src\2188752e3f-686fd8e62c.clean\extern\rg_etc1_v104 /nologo /DWIN32 /D_WINDOWS /W3 /utf-8 /GR /EHsc /MP  /MD /O2 /Oi /Gy /DNDEBUG /Z7  /O2 /Ob2 /Oi /Ot /Oy /GL -std:c++14 /showIncludes /Fosrc\nvtt\CMakeFiles\nvtt.dir\CompressionOptions.cpp.obj /Fdsrc\nvtt\CMakeFiles\nvtt.dir\nvtt.pdb /FS -c D:\buildtrees\nvtt\src\2188752e3f-686fd8e62c.clean\src\nvtt\CompressionOptions.cpp

[ras0219-msft]

Wow, that's a lot of bugs. Thanks!

[dg0yt]

@ras0219-msft Is it intentional that check_crt_linkage_of_libs is called for Build::LinkageType::STATIC, but not for Build::LinkageType::DYNAMIC in https://github.com/microsoft/vcpkg-tool/blob/main/src/vcpkg/postbuildlint.cpp#L1074-L1129?
I feel like the inconsistent linkage should have been discovered by the postbuild checks.