Change tinfoilhat data models

• solve issue OSTC/planning/ip-compliance#56 on TinfoilHat's side
  • implement path tags
  • separate git_sha1 and sha1_cksum fields for each source file
  • add variant field as short sha1 checksum of source files included in the recipe variant
• other data model adjustments and optimizations
  • move package_arch from recipe to package metadata
  • put depends_provides in a dict, with machine as key
  • use python dataclasses
  • remove unused metadata fields
• optimization for pipelines
  • collect package metadata from rpms (if available), in order to be able to work also with bb builds using shared sstate-cache dirs (rpms are saved in bb deploy dir even when build cache is used)
• test new models with real build dirs and verify the results