feat: nightly hardening - configuration validation guardrails

[mouse-value-add]

Problem

PromptLens accepted several invalid configuration values (e.g., out-of-range temperatures, non-positive token limits, empty model lists, unsupported output formats), which could cause runtime failures or silently skipped exports.

Approach

• Added strict Pydantic validators for:
  • ModelConfig.temperature (0..2)
  • ModelConfig.max_tokens (>0)
  • ExecutionConfig numeric bounds (parallel_requests, retry_attempts, retry_delay_seconds, timeout_seconds)
  • RunConfig.models non-empty requirement
• Hardened OutputConfig.formats by normalizing to lowercase, deduplicating, and rejecting unknown formats.
• Added focused test coverage in tests/test_config_validation.py.

Verification

• Ran: python3 -m pytest tests/test_config_validation.py
• Result: 5 tests passed.

Risks

• Existing configs that relied on invalid values (or unknown output format names) will now fail fast during validation.

Rollback Plan

• Revert this PR commit to restore previous permissive validation behavior.

[brainsparker]

Closing as superseded by #31, the most complete version of the config validation guardrails. It lands via the consolidation in #40.