feat(adf): inject configurable classifier/parsing rulesets

[SmartBrandStrategies]

Linked feedback: ADX-007

Problem

Classifier/parser heuristics are hardcoded, reducing reuse across organizations:

• fixed stay patterns and heading-to-module routing in classifier
• fixed imperative/advisory regex sets in markdown parser

Evidence:

• packages/adf/src/content-classifier.ts:53
• packages/adf/src/content-classifier.ts:76
• packages/adf/src/markdown-parser.ts:36

Scope

• Add optional config objects:
  • ClassifierConfig for stay patterns + heading/module routing + optional keyword rules
  • MarkdownStrengthConfig for imperative/advisory patterns
• Preserve existing behavior when config is omitted.
• Thread config through migration entry points (including CLI migrate path).

Acceptance Criteria

• Default behavior is unchanged and existing tests pass.
• New tests confirm custom rulesets alter classification/strength deterministically.
• CLI migrate can accept configured behavior without breaking current default UX.

Test Plan

• pnpm run test --filter @stackbilt/adf
• Add tests for:
  • custom stay pattern causing STAY
  • custom heading mapping to non-default module
  • custom strength patterns overriding defaults

[SmartBrandStrategies]

Implemented and merged in #21 (commit 1a9e547). Closing.