RFI flagging added to SDP Processing Function library

A central processing function library is a key component of the intended SKA architecture that addresses a number of NFRs. Establishing this library as soon as possible is important to reduce risk associated with understanding and providing the appropriate abstraction and interfaces to the library functions.
Adding usable processing functions as soon as the library is established enables their immediate use in the development of SDP workflows.

Porting RFI on GPUs should improve performance. To identify potential bottlenecks and areas which need closer investigation for improvement we need initial implementation of RFI sum-threshold algorithm.

Time-boxed activity to provide:

• RFI flagger processing function established in library (with appropriate release)
• Definition of done met
• Initial implementation of the RFI on GPU
• Asses bottlenecks of RFI implementation on GPUs on the confluence

The GPU version of the RFI flagger was implemented and tested against the CPU version of the RFI flagger. Different optimisations were explored, and those are summarised on a confluence page (https://confluence.skatelescope.org/display/SE/HIP-169%3A+Assessment+of+bottlenecks+of+GPU+implementation+of+sum+threshold+algorithm). GPU RFI flagger was also integrated into the processing function library (https://gitlab.com/ska-telescope/sdp/ska-sdp-func/-/tree/rfi_flagger).