AWS: Clean up DRM FDs before PCI rescan post axlf load

[karthdmg-xilinx]

Problem solved by the commit

AWS performs a pci rescan to change the PCI device ID during the first image load.
If there are any references to DRM render files which are not closed during the pci rescan,
XRT driver hangs waiting for ever to get the driver instance closed which usually happens if the DRM FDs are closed.

Bug / issue (if any) fixed, which PR introduced the bug, how it was discovered

XRT hangs indefinity in AWS environment during the PCI rescan after axlf load.

How problem was solved, alternative solutions (if any) and why they were rejected

Fix is to close all the opened DRM render FDs below the PCI rescan is issued.

Risks (if any) associated the changes in the commit

This is AWS specific change, would not impact other cards.

What has been tested and how, request additional testing if necessary

Ran RTL kernel APP to see the all the FDs are closed before PCI rescan, no hang observed.

Documentation impact (if any)

NA