Release Python GIL during slow load I/O

[andrew-bolin]

Problem solved by the commit

When using multiple FPGAs in multiple threads, loading XCLBINs cannot be performed concurrently as pyxrt holds the Python Global Interpreter Lock for the duration of the loading process.

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

See AMD WTS support case 00410749

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

GIL released during slow load operation. This fix was suggested by AMD support in November 2023, but was seemingly never submitted to the code base.

Risks (if any) associated the changes in the commit

Believed to be safe

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

AMD support said they have tested it. A Python snippet to demonstrate the original issue (and prove the fix) is available - https://gitlab.com/ska-telescope/low-cbf/ska-low-cbf-fpga/-/snippets/3613095

Documentation impact (if any)

None

[gbuildx]

andrew-bolin - is not a collaborator Can XRT admins please validate PR

[gbuildx]

Can one of the admins verify this patch?

[gbuildx]

andrew-bolin - is not a collaborator Can XRT admins please validate PR

[chvamshi-xilinx]

\build