Implement agnostic merger

[sdruskat]

• Implement a single merger class (working title ModelMerger) that merges 2 JSON-LD dictionaries (merge(source: expanded_json_ld, into: expanded_json_ld)).
• Make configurable with a strategy, via a config file (or, e.g., table in hermes.toml) and via the API.
• Refactor merge.py and path.py accordingly
• Strategy is an abstract class, of which instances map types/fields to treatments (requires Implement merging strategy API #339)
• Treatments provide logic to merge values of the given type/field (requires Implement fundamental "treatments" #340 )

This is a good case for using TDD.

[led02]

Being a child of the 80s having grown up with Intel assembler, op(dst, src) feels so much more natural. Also: does that mean source will be modified during the process?

[sdruskat]

Being a child of the 80s having grown up with Intel assembler, op(dst, src) feels so much more natural. Also: does that mean source will be modified during the process?

Point taken (although you're not that old), naming isn't final.

E.g., I think the class should not be called just Merger, but ModelMerger or similar while "treatments" should be called ValueMerger or similar, to differentiate the two levels of merging (complete models vs. single types/fields).