feat: comprehensive rate limit handling overhaul (inspired by opencode-antigravity-auth)

This commit addresses "Max retries exceeded" errors during stress testing where
all accounts would become exhausted simultaneously due to short per-second rate
limits triggering cascading failures.

## Rate Limit Parser (`rate-limit-parser.js`)
- Remove 2s buffer enforcement that caused cascading failures when API returned
  short reset times (200-600ms). Now adds 200ms buffer for sub-500ms resets
- Add `parseRateLimitReason()` for smart backoff based on error type:
  QUOTA_EXHAUSTED, RATE_LIMIT_EXCEEDED, MODEL_CAPACITY_EXHAUSTED, SERVER_ERROR

## Message/Streaming Handlers
- Add per-account+model rate limit state tracking with exponential backoff
- For short rate limits (< 1 second), wait and retry on same account instead
  of switching - prevents thundering herd when all accounts hit per-second limits
- Add throttle wait support for fallback modes (emergency/lastResort)
- Add `calculateSmartBackoff()` with progressive tiers by error type

## HybridStrategy (`hybrid-strategy.js`)
- Refactor `#getCandidates()` to return 4 fallback levels:
  - `normal`: All filters pass (health, tokens, quota)
  - `quota`: Bypass critical quota check
  - `emergency`: Bypass health check when ALL accounts unhealthy
  - `lastResort`: Bypass BOTH health AND token bucket checks
- Add throttle wait times: 500ms for lastResort, 250ms for emergency
- Fix LRU calculation to use seconds (matches opencode-antigravity-auth)

## Health Tracker
- Increase `recoveryPerHour` from 2 to 10 for faster recovery (1 hour vs 5 hours)

## Account Manager
- Add consecutive failure tracking: `getConsecutiveFailures()`,
  `incrementConsecutiveFailures()`, `resetConsecutiveFailures()`
- Add cooldown mechanism separate from rate limits with `CooldownReason`
- Reset consecutive failures on successful request

## Base Strategy
- Add `isAccountCoolingDown()` check in `isAccountUsable()`

## Constants
- Replace fixed `CAPACITY_RETRY_DELAY_MS` with progressive `CAPACITY_BACKOFF_TIERS_MS`
- Add `BACKOFF_BY_ERROR_TYPE` for smart backoff
- Add `QUOTA_EXHAUSTED_BACKOFF_TIERS_MS` for progressive quota backoff
- Add `MIN_BACKOFF_MS` floor to prevent "Available in 0s" loops
- Increase `MAX_CAPACITY_RETRIES` from 3 to 5
- Reduce `RATE_LIMIT_DEDUP_WINDOW_MS` from 5s to 2s

## Frontend
- Remove `capacityRetryDelayMs` config (replaced by progressive tiers)
- Update default `maxCapacityRetries` display from 3 to 5

## Testing
- Add `tests/stress-test.cjs` for concurrent request stress testing

Co-Authored-By: Claude <noreply@anthropic.com>

src/constants.js

@@ -103,16 +103,33 @@ export const MAX_ACCOUNTS = config?.maxAccounts || 10; // From config or 10
 // Rate limit wait thresholds
 export const MAX_WAIT_BEFORE_ERROR_MS = config?.maxWaitBeforeErrorMs || 120000; // From config or 2 minutes
 
-// Gap 1: Retry deduplication - prevents thundering herd on concurrent rate limits
-export const RATE_LIMIT_DEDUP_WINDOW_MS = config?.rateLimitDedupWindowMs || 5000; // 5 seconds
+// Retry deduplication - prevents thundering herd on concurrent rate limits
+export const RATE_LIMIT_DEDUP_WINDOW_MS = config?.rateLimitDedupWindowMs || 2000; // 2 seconds
+export const RATE_LIMIT_STATE_RESET_MS = config?.rateLimitStateResetMs || 120000; // 2 minutes - reset consecutive429 after inactivity
+export const FIRST_RETRY_DELAY_MS = config?.firstRetryDelayMs || 1000; // Quick 1s retry on first 429
+export const SWITCH_ACCOUNT_DELAY_MS = config?.switchAccountDelayMs || 5000; // Delay before switching accounts
 
-// Gap 2: Consecutive failure tracking - extended cooldown after repeated failures
+// Consecutive failure tracking - extended cooldown after repeated failures
 export const MAX_CONSECUTIVE_FAILURES = config?.maxConsecutiveFailures || 3;
 export const EXTENDED_COOLDOWN_MS = config?.extendedCooldownMs || 60000; // 1 minute
 
-// Gap 4: Capacity exhaustion - shorter retry for model capacity issues (not quota)
-export const CAPACITY_RETRY_DELAY_MS = config?.capacityRetryDelayMs || 2000; // 2 seconds
-export const MAX_CAPACITY_RETRIES = config?.maxCapacityRetries || 3;
+// Capacity exhaustion - progressive backoff tiers for model capacity issues
+export const CAPACITY_BACKOFF_TIERS_MS = config?.capacityBackoffTiersMs || [5000, 10000, 20000, 30000, 60000];
+export const MAX_CAPACITY_RETRIES = config?.maxCapacityRetries || 5;
+
+// Smart backoff by error type
+export const BACKOFF_BY_ERROR_TYPE = {
+    RATE_LIMIT_EXCEEDED: 30000,      // 30 seconds
+    MODEL_CAPACITY_EXHAUSTED: 15000, // 15 seconds
+    SERVER_ERROR: 20000,             // 20 seconds
+    UNKNOWN: 60000                   // 1 minute
+};
+
+// Progressive backoff tiers for QUOTA_EXHAUSTED (60s, 5m, 30m, 2h)
+export const QUOTA_EXHAUSTED_BACKOFF_TIERS_MS = [60000, 300000, 1800000, 7200000];
+
+// Minimum backoff floor to prevent "Available in 0s" loops (matches opencode-antigravity-auth)
+export const MIN_BACKOFF_MS = 2000;
 
 // Thinking model constants
 export const MIN_SIGNATURE_LENGTH = 50; // Minimum valid thinking signature length
@@ -258,10 +275,16 @@ export default {
     MAX_ACCOUNTS,
     MAX_WAIT_BEFORE_ERROR_MS,
     RATE_LIMIT_DEDUP_WINDOW_MS,
+    RATE_LIMIT_STATE_RESET_MS,
+    FIRST_RETRY_DELAY_MS,
+    SWITCH_ACCOUNT_DELAY_MS,
     MAX_CONSECUTIVE_FAILURES,
     EXTENDED_COOLDOWN_MS,
-    CAPACITY_RETRY_DELAY_MS,
+    CAPACITY_BACKOFF_TIERS_MS,
     MAX_CAPACITY_RETRIES,
+    BACKOFF_BY_ERROR_TYPE,
+    QUOTA_EXHAUSTED_BACKOFF_TIERS_MS,
+    MIN_BACKOFF_MS,
     MIN_SIGNATURE_LENGTH,
     GEMINI_MAX_OUTPUT_TOKENS,
     GEMINI_SKIP_SIGNATURE,