this can avoid sooo many reinterpret_cast<> in application code

when chunk size was a already multiple of the page size, the
resulting size would be one page too much

Now supports 64 words per frame

This was neccessary in order to make the memory available via GDRcopy
when multiple small allocations were made. cudaMalloc() would return
multiple memory chunks located in the same GPU page, which GDRcopy
pretty much dislikes (`gdrdrv:offset != 0 is not supported`).
As a side effect, this will keep the number of BAR-mappings done
via GDRcopy low, because they seem to be quite limited.

This is useful when we sub-delegate management of a memory block
to another allocator.

Virtual inheritance is required because (for example) the Rtds2Gpu
IP inherits from Hls and IpNode who both inherit from IpCore.

 - better tested IP (testbenches)
 - detect invalid frame sizes
 - more status reporting