Add model-independent LLM provider abstraction with schema validation

feat(ui): batch risk column + bulk vault controls; isolate test vaults

feat(ats): AUTO_AGGRESSIVE mode + risk severity levels

feat(ats): independent validator + anti-cheat — break circular self-scoring

feat(ats): Candidate Experience Vault + UI status tiers (90%+ pipeline complete)

feat(ats): v2 pipeline — structured JD, evidence matching, weighted 2-part score

fix(ats): honest meaningful JD-driven keywords — no stuffing, no inflation

fix(phase4.4): allowlist keyword extraction — the real root-cause fix

fix(phase4.3): systemic JD keyword extraction — works on any new JD

fix(phase4.2): backfill dropped roles + enforce recruiter pitch

fix(phase4.1): aggressive bullet weaving + canonical-tuned scoring

feat(phase4): canonical Resume model + single locked visual format

fix(ui+ats): remove title, fix dark-on-dark inputs, aggressive injection for 90%+

feat(ats): Phase 3 — lemma+phrase scorer + liberal LLM policy + recruiter pitch