feat: add rate limit handling with automatic retry for OpenRouter API

[armand0e]

Add global rate limit state tracking per model/endpoint and implement retry logic with exponential backoff for 429 responses. Parse x-ratelimit-reset and x-ratelimit-remaining headers to coordinate wait times across concurrent requests. Retry up to 5 times with calculated delays based on reset timestamps or fallback to exponential backoff (1s, 2s, 3s, etc.) in case outputs aren't successfully parsed. Non-invasive and non-breaking.

All tests passed:

✔ parseArgs requires model and prompts (1.0002ms)
✔ parseArgs defaults store-system to true (0.0923ms)
✔ parseArgs defaults concurrent to 1 (0.0497ms)
✔ parseArgs parses --concurrent (0.065ms)
✔ parseArgs parses OpenRouter provider flags (1.4609ms)
✔ parseArgs parses --reasoningEffort (0.079ms)
✔ parseArgs parses --openrouter.isFree (0.0592ms)
✔ parseArgs supports --config YAML (10.1642ms)
✔ parseArgs lets CLI override config (4.4964ms)
✔ buildRequestMessages omits system when empty (0.1571ms)
✔ buildOutputMessages respects storeSystem flag (0.1172ms)
✔ formatAssistantContent wraps reasoning in <think> (0.0488ms)
✔ callOpenRouter sends correct payload and parses reasoning (0.3448ms)
✔ callOpenRouter includes provider prefs when provided (0.1282ms)
✔ callOpenRouter includes reasoning.effort when provided (0.0767ms)
✔ callOpenRouter reasoning.effort works for non-OpenRouter apiBase (0.0646ms)
✔ ensureReadableFile throws if missing or not a file (1.9111ms)
ℹ tests 17
ℹ suites 0
ℹ pass 17
ℹ fail 0
ℹ cancelled 0
ℹ skipped 0
ℹ todo 0
ℹ duration_ms 75.0296

[armand0e]

dont merge yet. I will be bulletproofing this later today to ensure this is the final PR for this functionality