nvenc improvements: fast resize / reconfigure, rgb modes, max grid sizes, etc

[totaam]

Split from #466, see also #564:

• handle resize without re-init? (don't think so, I think this would require client side support)
• handle reconfiguration: swapping NV12 mode for YUV444P on the fly with speed/quality changes
• maybe do parallel upload with encode for YUV444P mode? (save latency)
• upload pixels in place? (skip inputBuffer)
• deal with GPU suspend-resume cycles (see r5110 for opencl in opencl acceleration for csc and/or encoding #422) - here is the CUDA error with nvenc:

  File "encoder.pyx", line 1588, in xpra.codecs.nvenc.encoder.Encoder.compress_image (xpra/codecs/nvenc/encoder.c:12085)
  File "encoder.pyx", line 1624, in xpra.codecs.nvenc.encoder.Encoder.do_compress_image (xpra/codecs/nvenc/encoder.c:12598)
LogicError: cuMemcpyHtoD failed: invalid/unknown error code

Minor / easy:

• honouring max_block_sizes, max_grid_sizes and max_threads_per_block - doesn't seem to be causing problems yet
• handle other RGB modes in kernel (easy - allows us to run in big endian servers)
• try using nvenc on win32 for shadow servers
• access nvenc encoder statistics info?

[totaam]

See also #3518 and #3286