Bugfix: improved check_deltakpoly in sicdcheck

[pressler-vsc]

Description

This PR:

• improves the method for determining the expected DeltaKs in sicdcheck's check_deltakpoly
  • currently, only the ValidData vertices are used (or FullImage corners if missing). In addition to being rather sparse, these are possibly outside of the extent of the SICD in question
  • this branch proposes an adapted version of the method recently added to sarkit-processing
• improves the docstrings/messages and downgrades the check from need -> want
  • because the standard doesn't explicitly prescribe how these should be filled out; I believe want is more appropriate