Poll controllers on a separate thread on Windows

[unknownbrackets]

This means controllers will sample more than the game fps, and will work whenever keyboards do.

But, consider a game that is running at 30fps:

• The cpu runs as fast as it can for 1 frame (1/30 of an emulated second.)
• The cpu sleeps to wait for the next frame.
• The user presses buttons during the 1/30 seconds that the game might respond to (per vblank or delay thread, etc.)
• Nothing happens until the next frame's cpu burst.

Which means, the faster the device, the worse the input latency will be. This doesn't change that, but at least if the game isn't flipping, it will ensure that controller inputs still work just as well as keyboard inputs.

Maybe we should have an option for faster devices that allows you to force the cpu to checkpoint at least each vblank, or possibly more frequently. I think that would actually help #4020 (assuming the input device provides enough accuracy.)

-[Unknown]

[solarmystic]

@unknownbrackets
I had a quick go with two rhythm based games (Project Diva 2nd and Diva Extend) after applying this commit to the current master.

I don't know how to describe it exactly, but the button presses in those games definitely feels a bit more responsive as compared to the current master. My scores were definitely improved, although that could just be a placebo effect of the commit; I was of course expecting an improvement. The situation (for me) prior to this commit wasn't as dire as others have reported however, I did finish both games on Hard Difficulty before this on PPSSPP.

Tested on my Windows 7 laptop with a Microsoft X360 wired controller.

[unknownbrackets]

It may help but I'm not expecting it to fix the issue, especially for faster devices as I described.

-[Unknown]

[solarmystic]

PPSSPP crashes after exiting the program with the initial unknownbrackets@a78d38c commit . Still does after updating it to unknownbrackets@5c707f8

Stack trace from MSVC 2013:-

    ntdll.dll!00000000772512f7()    Unknown
    KernelBase.dll!000007fefd161873()   Unknown
    kernel32.dll!0000000077002f51() Unknown
>   PPSSPPWindows64.exe!InputDevice::StopPolling() Line 83  C++
    PPSSPPWindows64.exe!WinMain(HINSTANCE__ * _hInstance, HINSTANCE__ * hPrevInstance, char * szCmdLine, int iCmdShow) Line 438 C++
    [External Code] 

[unknownbrackets]

Oh, I forgot to detach. It wasn't crashing for me (except in that weird race condition where host had been deleted inside NativeShutdown()), but detaching may fix it for you and should be correct.

-[Unknown]

[solarmystic]

Yep, unknownbrackets@b641ced did the trick, thanks. Should be good to go now.