[CI] Move Windows builds away from hardcoded -j2

[svenvh]

Instead of the hardcoded -j2, use the build tool's default number.

Fixes #3481

[spencer-lunarg]

hmmm... seem this didn't improve the build time at all... still "seems correct", but was hoping for more... I blame MSVC 😆

[svenvh]

Trying another approach: $env:NUMBER_OF_PROCESSORS

[spencer-lunarg]

... @svenvh i'm invested in this, I have an windows laptop, will pull it out next week and try and see if can figure out locally how to make it faster in CI (I can also just test in a fork, normally when I test this stuff, I push my branch up and it runs in my fork, but I also just erase all the other jobs so it only has to run one at time)

[svenvh]

... @svenvh i'm invested in this, I have an windows laptop, will pull it out next week and try and see if can figure out locally how to make it faster in CI (I can also just test in a fork, normally when I test this stuff, I push my branch up and it runs in my fork, but I also just erase all the other jobs so it only has to run one at time)

Awesome, thanks! It seems $NUMBER_OF_PROCESSORS does not improve matters so I'm out of ideas, more than happy for you to give it a try.

[spencer-lunarg]

experimented in my branch/fork, going --parallel didn't seem to help

I then tried forcing the cmake with

if(MSVC)
    add_compile_options(/MP)
endif()

and still 50min to build... so seems the build issue on windows is not from number of cores used, but likely something deeper with how it is being built or maybe MSVC is just having a harder time here