fix(eval): reject per-case target selection - #1551

Summary

Authored eval YAML no longer supports target selection inside individual test-case execution blocks. The test-case execution schema now omits target and targets, and the parser/validator reject unsupported test execution keys through generic strictness instead of a bespoke field-specific deprecation path.

Runtime matrix execution no longer reads EvalTest.targets, so every selected suite/experiment/CLI target runs the same filtered case set rather than silently creating sparse per-case matrices. Promptfoo imports keep default provider filters at the suite target level and reject test-level provider filters because they would require unsupported per-case target selection.

Docs, examples, the eval-writer guidance, and the generated schema reference now show target choice as runtime policy rather than test-case data.

Validation

• bun install --frozen-lockfile (dependencies were absent in this isolated worktree; lockfile unchanged)
• bun --filter @agentv/core build
• bun test packages/core/test/evaluation/matrix-targets.test.ts packages/core/test/evaluation/eval-inline-experiment.test.ts packages/core/test/evaluation/validation/eval-file-schema.test.ts packages/core/test/evaluation/validation/eval-validator.test.ts packages/core/test/evaluation/loaders/config-loader.test.ts apps/cli/src/commands/import/promptfoo.test.ts apps/cli/test/commands/eval/task-bundle.test.ts packages/core/test/evaluation/validation/eval-schema-sync.test.ts
• bun run typecheck
• bun run lint
• git diff --check

Post-Deploy Monitoring & Validation

No additional operational monitoring required. This is local parser/validator/CLI behavior with no deployed runtime service; GitHub Actions are the merge gate, and failures in schema sync, core parser tests, CLI import tests, typecheck, or lint should block release.