@:posInfos to pass haxe.PosInfos to Reflect.makeVarArgs()

I think one of the most common use-cases for `Reflect.makeVarArgs()` would be for logging tools, and in the same breath one of the most common use-cases for `haxe.PosInfos` would be the usage in debug tools and loggers as well, since we can read out the line numbers of the call.

However those 2 don't work well together, although you would expect that that's how `trace()` works under the hood :
- It can take any number of arguments
- The last argument is `PosInfos`

Now currently if you want to replicate that behaviour, the only way is to redefine `trace()`. But in html5, you might not want to do that, since you don't want to just log out to `console.log` - you want to log out to `console.warn`, `console.error`, `console.debug`, `console.info` etc. - and of course you want to attach the line number of the call to it.

So assuming there's no other way to have a function with an arbitrary amount of arguments + `haxe.PosInfos` attached to it ( as it is in `trace` ), I would propose to add a `@:posInfos` metadata, that passes on `haxe.PosInfos` as the last parameter to a function when it's called.

This would also make the usage of `trace()` more standard and understandable, and would help with alternative logger solutions without them all having to redefine `trace` and possibly colliding with each other.

If I've missed anything and there is actually another alternative solution please feel free to play around with this example : https://try.haxe.org/#361D7

Start with the linked try.haxe.org example and inspect the current behavior of Reflect.makeVarArgs() and trace(). Determine whether a metadata-based way to pass haxe.PosInfos is viable, and define completion as supporting arbitrary arguments with PosInfos attached for alternative logger functions.

Written by the indexing model from the issue text.