Make GLM include directory portable.

[Robadob]

• pyflamegpu.GLM now reports GLM status
  • FLAMEGPU_USE_GLM is the pre-existing C macro
• When searching for GLM include files, RTC now tries the GLM_INC_DIR environment variable
• Package GLM include inside wheel alongside FLAMEGPU headers
• When importing pyflamegpu, GLM_INC_DIR is set to the location of GLM include dir within the module's install

• Package GLM headers/license inside python wheel
• Setup GLM_INC_DIR environment variable to be used
• Document GLM_INC_DIR environment variable
• Add GLM flag to SWIG
• Only setup GLM_INC_DIR inside wheel when GLM enabled in build

Closes #1001

[Robadob]

All tests ran using --gtest_filter=*glm*, With Windows/Debug/GLM=ON

C, Without using environment var

[----------] Global test environment tear-down
[==========] 34 tests from 24 test suites ran. (270287 ms total)
[  PASSED  ] 34 tests.

C, after deleting tmp/flamegpu/jitifycache and SET GLM_INC_DIR=c:/windows.

[----------] 2 tests from DeviceRTCAPITest
[ RUN      ] DeviceRTCAPITest.AgentFunction_array_get_glm
C:\Users\Robert\fgpu2\fgpu2\src\flamegpu\detail\JitifyCache.cu(235): Error environment variable GLM_INC_DIR (c:/windows) does not contain glm/glm.hpp. Please correct this environment variable.

C, after SET GLM_INC_DIR=C:\Users\Robert\fgpu2\fgpu2\build\_deps\glm-src2 (renamed dir to be sure it wasn't cheating).

[----------] Global test environment tear-down
[==========] 34 tests from 24 test suites ran. (271615 ms total)
[  PASSED  ] 34 tests.

[Robadob]

Can't find any GLM tests in python test suite (would explain pyflamegpu.GLM) not being a thing.

Doesn't appear SWIG has GLM mappings setup either (did this get left on hold because of the Jitify slow compilation?). Regardless, the current changes should be appropriate, as it can still be used within RTC without being directly used in Python.

[ptheywood]

Needs atleast one python test which uses GLM (if GLM is enabled. Call pytest.skip(message) in the test(s) if pyflamegpu.GLM is not enabled)

Env var should be prefixed with FLAMEGPU_GLM_INC_DIR (unless GLM_INC_DIR is an env var set by GLM if centrally installed, which I doubt, and even then it needs to match the version used to build flamegpu, so should be a flamegpu specific one).

Not sure the warning on module import is a good addition either. importing a python module should be quiet IMO.

[Robadob]

Not sure the warning on module import is a good addition either. importing a python module should be quiet IMO.

Line 6 of the same file, there's already a warning for the FLAMEGPU inc dir version, I just copied similar pattern.

Otherwise good suggestions, will apply later in the week when I'm back on proper FGPU stuff. GLM + py is still a bit experimental.

[ptheywood]

Line 6 of the same file, there's already a warning for the FLAMEGPU inc dir version, I just copied similar pattern.

That's fair (and I'd prolly lean towards getting rid of that too to be honest, but can be a future thing)

[ptheywood]

Fixed a linux warning from ~1 year ago in a glm-only test, but otherwise looks good and tests pass for me.