We create three buffers to store data for use with MTLCommandBuffer.

During performance testing, we observed a scenario where all three command buffers could be submitted for execution but remain incomplete. In such cases, we end up waiting for all three command buffers to finish, leading to a longer wait time and causing the CPU to stall while encoding commands.

To optimize this, we need to modify the approach so that we wait for only one MTLCommandBuffer to complete execution, which would release a buffer.