fix(cli): add evaluation cache to miner score stub

[statxc]

Summary

gitt miner score runs the validator scoring pipeline through _StubValidator, which avoids the real validator runtime dependencies. After #1040, get_rewards() always calls self.evaluation_cache.evict_many(penalized_uids), but the CLI stub did not define evaluation_cache.

This PR adds a lazy MinerEvaluationCache factory and wires it onto _StubValidator, matching the validator surface now required by get_rewards() while keeping CLI imports lightweight.

Changes

• Add _new_evaluation_cache() to lazily construct MinerEvaluationCache.
• Attach self.evaluation_cache in _StubValidator.__init__.
• Add regression coverage that exercises the real oss_contributions -> get_rewards path instead of mocking oss_contributions away.

Testing

.venv/bin/python -m pytest tests/cli/test_miner_score.py tests/validator/test_validator_cache_fallback.py -q
.venv/bin/python -m pytest tests/cli -q
.venv/bin/python -m ruff check gittensor/cli/miner_commands/score.py tests/cli/test_miner_score.py tests/validator/test_validator_cache_fallback.py
.venv/bin/python -m ruff format --check gittensor/cli/miner_commands/score.py tests/cli/test_miner_score.py
.venv/bin/python -m compileall -q gittensor/cli/miner_commands/score.py tests/cli/test_miner_score.py
git diff --check
UV_CACHE_DIR=/tmp/uv-cache .venv/bin/python -m pre_commit run --files gittensor/cli/miner_commands/score.py tests/cli/test_miner_score.py

[anderdc]

Take the guard at the call site instead. In gittensor/validator/oss_contributions/reward.py:143, wrap the eviction in if penalized_uids: — the single-miner dev flow never produces penalized UIDs, so the attribute is never accessed and the stub stays minimal. Drop the _new_evaluation_cache factory and _StubValidator.evaluation_cache change from this PR. Keep the new test — that lower-mock shape (patching only evaluate_miners_pull_requests) is the right coverage gap to close.

[statxc]

Hi @anderdc how are you? I've solved this correctly. Thanks for reviewing.
Could you add bug label on this PR? In my opinion, this PR means bug fix too.

[statxc]

Thanks @anderdc