Windows impl. still pending.

This appears to be broken. Why should a table where all rows want to be kept
together be moved one page forward just because there is something
before it on the page where it is? It either fits its curent page, in which
case it's wrong, or it does not fit, in which case it may not fit the following
page either and something else needs to take care of it anyway.

A temporary tweak to restore build-ability; the oslDirectory
change needs a little more thought & public exposure, but rsc
no longer crashes during compile ...

 - allow to specify view boxes for sub paths

Replaced calls to operator>>(long) with operator>>(sal_Int32)
Replaced calls to operator<<(long) with sal::static_int_cast<sal_Int32>

Replaced calls to operator<<(long) with sal::static_int_cast<sal_Int32>
Replaced calls to operator>>(long) with operator>>(sal_Int32)

Replaced calls to operator<<(long) with sal::static_int_cast<sal_Int32>
Replaced calls to operator>>(long) with operator>>(sal_Int32)

Replaced calls to operator<<(long) with sal::static_int_cast<sal_Int32>
Replaced calls to operator>>(long) with operator>>(sal_Int32)

Changed case SbxSALINT64: to use operator>>(sal_Int64)

Replaced calls to operator<<(long) with sal::static_int_cast<sal_Int32>
Replaced calls to operator>>(long) to use sal_Int32

Removed declarations & definitions for operator<<(long),(int)&(short)
Removed declarations & definitions for operator>>(long),(int)&(short)
Added (where necessary) operator<< for sal_Int & sal_uInt types
Added (where necessary) operator>> for sal_Int & sal_uInt types
Added SwapInt64 function, basically a copy of SwapUInt64

Part L
Modules
ucb

vcl's DrawBitmap methods have a very nice optimisation, that reduce
the blit to exactly the area covered (i.e. taking into account clip,
scale etc.). This was hand-coded, now moved out to own AdjustTwoRect
method, and also used from DrawOutDev / CopyArea.

No classes derived from this and they are not virtual methods.