feat: Refactoring of rules configuration and rules converter

[kentquirk]

Which problem is this PR solving?

This is a reworking of the refinery rules structure. The existing rules configuration format is both awkwardly structured for technical reasons, as well as just being confusing. People make a lot of mistakes.

This structure organizes refinery rules into a map of environment name (dataset in classic mode) to sampler config, and reserves the name __default__ for the default case. It also puts the sampler configurations indented under the sampler type, which avoids another level of confusion.

Short description of the changes

• Refactor the sampler_config file so all the configuration types are near each other
• Rename the Rule and Condition slices to Rules and Conditions
• Decorate all the types with yaml (and sometimes json) tags
• Add a "rules" command to the conversion tool that uses this stuff to do the conversion

This work isn't "done" (there is more planned) but it's stable enough to call it a checkpoint, review as-is, and do the next things as separate PRs.