Fix DXGI present failures caused by race conditions

[praydog]

In some scenarios (seen while loading or particularly high, uncapped framerate scenarios), resetting the command allocator before the previous command list had finished executing on the GPU would cause DXGI present to fail with DXGI_ERROR_DEVICE_HUNG or some other similar error, and the game would subsequently crash.

Seen 100% during loading screens with uncapped, variable framerate in Dragon's Dogma 2. This could've also happened sometimes when the backbuffer was resized (aka on_device_reset), then reframework-d2d would release all of its resources sometimes before the command lists using them were finished.

This PR does the following things:

• Splits the original command allocator and command list into multiple command contexts with their own fences.
• Likewise, splits the vertex buffers into the same amount of command queues allocated to prevent conflicting access of the GPU buffer. This part might not have been necessary if it's just a constant vertex buffer.

CommandContext.hpp/.cpp are slimmed down versions of the ones used in REFramework/UEVR/FF7R-UEVR.

This could arguably be much simpler if 1 fence was just used for the original command list/allocator, however this has the drawback of lowering performance in double/triple-buffered scenarios (as is common on D3D12 with multiple backbuffers)