[brl.reflection] 32 bit build issues with LONG-params

[GWRon]

I received a bug report by a user ... somehow a lua script which invokes some BlitzMax-exposed method received wrong values.

I am not able to replicate it on my computers (linux, windows7, windows10, AMD and Intel cpus).

If building a 64bit binary for him, the problem is gone. It only happens for 32bit builds. That user had other special "results" when I used "long" in c code rather than "BBLONG" (fixed thanks to @HurryStarfish ).

The method is defined as Method GetDay:Int(useTime:Long = -1) - so the parameter is a LONG. I call the method via reflection.
Before invoking the method I print out the "to pass" parameters. Inside the method I print out what is received - and what is returned at the end. After invokation is done I print out what was returned.

Lua -> BlitzMax works. So param is read as it should.
The invoked method receives an incorrect value .. .and calculates with it.
It returns a value (which it printed out).
The value the invocation result contains is different to the printed out one.

I cannot write up a simple example but something like this can help:

	'attention: LUA uses a default param of "0"
	'-> so for this and other functions we have to use "<=0" instead of "<0"
	Method GetDay:Int(useTime:Double = -1) {_exposeToLua}
		print "useTime="+useTime
		Return useTime * 10
	End Method

[GWRon]

Replacing the parameter with other types makes it work (I had it on "double" before rewriting my stuff to use longs for times not doubles).

[GWRon]

As stated in #227 issues arise also for "doubles" - if you do debug builds instead of release builds.