APR-57: add teacher model discovery flag

[AmeenP]

Summary

• add a typed Hosted Training client method for GET /rft/teacher-models
• add prime train models --teachers to list generate-capable teacher inference models for OPD/SFT-style teacher configs
• keep default prime train models unchanged because it lists trainable student base models and capacity, not teacher scoring candidates

Stack

platform#2342 pi-inference execution
  POST /api/v1/generate
  POST /api/inference/v1/generate
          |
          v
platform#3007 pi-inference discovery
  GET /api/internal/models -> capabilities.generate
          |
          v
platform#3008 RFT consumption
  GET /api/v1/rft/teacher-models
          |
          v
prime#699 OPD config surface
  loss = "opd" + [teacher]
          |
          v
this PR CLI discovery
  prime train models --teachers

• Stacked on APR-57: add hosted OPD CLI config #699 (feat/hosted-opd-cli).
• Runtime/API dependency is platform#3008, which is itself stacked on platform#3007 and platform#2342.

Validation

• .venv/bin/ruff format packages/prime/src/prime_cli/api/rl.py packages/prime/src/prime_cli/commands/rl.py packages/prime/tests/test_rl_api.py packages/prime/tests/test_rl_models.py
• .venv/bin/ruff check packages/prime/src/prime_cli/api/rl.py packages/prime/src/prime_cli/commands/rl.py packages/prime/tests/test_rl_api.py packages/prime/tests/test_rl_models.py
• .venv/bin/python -m pytest packages/prime/tests/test_rl_api.py packages/prime/tests/test_rl_models.py (20 passed)
• git diff --check

Linear: APR-57