This is a huge one, sorry for that.

These commits get rid of raw platform/process/library/compilers defines, i.e. _MSC_VER. And it replaces MSVC warning pragmas with EA_DISABLE_VC_WARNING and the likes. I think this is more clear, and easier to read/reason about.

I can split the first commit up in smaller, easier digestible commits if needed.

Only require EABase to be recursively (git submodule) cloned if EABase target isn't already present.

Solved in the same manner as https://github.com/electronicarts/EASTL/blob/master/CMakeLists.txt#L49

When trying to build EAThread using the 64 bit compilers from the NDK, I get the following errors:

In file included from ../../library/eaoss/src/EAThread/eathread.cpp:89:
../../library/eaoss/src/EAThread/android/eathread_fake_atomic_64.cpp:37:26: error: cast from pointer to smaller type 'unsigned int' loses information
        pthread_mutex_t* lock = EAT_SWAP_LOCK(addr);
                                ^~~~~~~~~~~~~~~~~~~
../../library/eaoss/src/EAThread/android/eathread_fake_atomic_64.cpp:31:54: note: expanded from macro 'EAT_SWAP_LOCK'
#define EAT_SWAP_LOCK(addr) &sFakeAtomic64SwapLocks[((unsigned)(void*)(addr) >> 3U) % EAT_FAKE_ATOMIC_SWAP_LOCK_COUNT]
                                                     ^~~~~~~~~~~~~~~~~~~~~~~
../../library/eaoss/src/EAThread/android/eathread_fake_atomic_64.cpp:52:26: error: cast from pointer to smaller type 'unsigned int' loses information
        pthread_mutex_t* lock = EAT_SWAP_LOCK(addr);
                                ^~~~~~~~~~~~~~~~~~~
../../library/eaoss/src/EAThread/android/eathread_fake_atomic_64.cpp:31:54: note: expanded from macro 'EAT_SWAP_LOCK'
#define EAT_SWAP_LOCK(addr) &sFakeAtomic64SwapLocks[((unsigned)(void*)(addr) >> 3U) % EAT_FAKE_ATOMIC_SWAP_LOCK_COUNT]
                                                     ^~~~~~~~~~~~~~~~~~~~~~~
../../library/eaoss/src/EAThread/android/eathread_fake_atomic_64.cpp:73:26: error: cast from pointer to smaller type 'unsigned int' loses information
        pthread_mutex_t* lock = EAT_SWAP_LOCK(addr);
                                ^~~~~~~~~~~~~~~~~~~
../../library/eaoss/src/EAThread/android/eathread_fake_atomic_64.cpp:31:54: note: expanded from macro 'EAT_SWAP_LOCK'
#define EAT_SWAP_LOCK(addr) &sFakeAtomic64SwapLocks[((unsigned)(void*)(addr) >> 3U) % EAT_FAKE_ATOMIC_SWAP_LOCK_COUNT]
                                                     ^~~~~~~~~~~~~~~~~~~~~~~
3 errors generated.

Changing that macro to use unsigned long for 64 bit builds eliminates the error, but I'm not sure if that will break functionality since it's not clear to me what that macro is doing.

Change EABase from link to include - the libraries won't build otherwise on Ubuntu. I'm guessing that in the past EABase was a binary library to link against, and was changed to header only - but I didn't chase its history.

In this PR I made MinGW-w64 compilers work again with EAThread.
I opted for using pthreads, as those are native to MinGW-w64 and GCC has no support for SEH intrinsics, such as __try and __catch.

I tested this with GCC, Clang, MSVC, and Clang-CL, but I can't remember if I tested it on Linux.

If possible I'd like to extend EAThread to be numa-aware somewhere in the future, so that MinGW users can make use of all threads on new high core count CPUs. I'm assuming the MSVC side is already numa aware.

As usual, let me know if there are any issues with the PR :)

eathread_pthread_stack_info.cpp in both Kettle and unix folders causes the MSVC compiler to complain.

Renaming them to _kettle and _unix removes the warning.