Added timestamp to KeyDown and KeyUp events.

[EliteMasterEric]

I am looking into methods of retrieving a highly precise timestamp corresponding with a user's keypress. This is important for applications which need to know exactly when a given key was pressed (which, among other things, includes video games such as rhythm games or fighting games, which demand precise inputs).

If each key event includes a precise hardware timestamp, recorded before any of the processing required to call event listeners is done, then said processing can be compensated for by retrieving a new timestamp and calculating the difference.

I discovered in my research that the SDL_KeyboardEvent which Lime receives (on native) does include such a timestamp, however the value is not retained, so I made this pull request to include it.

The pull request is a draft pending the resolution of the following issues:

• The change to the method signature of lime.app.Application.onKeyDown is a breaking change which affects OpenFL, Flixel, and other dependent libraries, is there a more stable method to make this change?
• The value provided in the SDL_KeyboardEvent is millisecond accurate, and is the value of SDL_GetTicks function, which is based on the duration since SDL was initialized, but there is not currently an accessible method to retrieve this value in Lime to perform comparison with.
• HTML5 platforms also have a timestamp associated with keyboard events, but the timestamp value is an absolute epoch time instead. How should this discrepancy be handled?

[player-03]

The change to the method signature of lime.app.Application.onKeyDown is a breaking change which affects OpenFL, Flixel, and other dependent libraries, is there a more stable method to make this change?

For now we could use a compile-time define, like #if lime_key_timestamps. Change the method signature if and only if they set the define, and then users who don't opt in won't experience any changes.

HTML5 platforms also have a timestamp associated with keyboard events, but the timestamp value is an absolute epoch time instead. How should this discrepancy be handled?

This has precedent in the Timer.stamp() function, so we could handle it the same way. Simply tell people that they should only look at differences between values, not at the individual values themselves.

The value provided in the SDL_KeyboardEvent is millisecond accurate, and is the value of SDL_GetTicks function, which is based on the duration since SDL was initialized, but there is not currently an accessible method to retrieve this value in Lime to perform comparison with.

I wonder if this change would be better off as its own Haxelib. Not saying it would, necessarily, just that it's something to consider.

Pros:

• Whenever you found a new function you need to expose, you could just do it.
• You could focus on C++ in particular, instead of worrying about cross-platform consistency.
• You wouldn't have to worry about breaking changes.

Cons:

• It wouldn't be integrated into OpenFL/Lime, though that seems manageable. You didn't need these events to be tied to the display list, did you?