fix: use task stored API config as fallback for rate limit

When provider state is unavailable (state is undefined/null), fall back to the task's stored apiConfiguration for rate limiting instead of defaulting to 0.

This prevents rate limiting from being silently disabled in edge cases where getState() fails or provider reference is temporarily lost.

Changes:
- Line 3643 (attemptApiRequest): Use (apiConfiguration ?? this.apiConfiguration)
- Line 3974 (backoffAndAnnounce): Use (state?.apiConfiguration ?? this.apiConfiguration)

Both locations now follow the same pattern:
1. Prefer the most up-to-date config from provider state
2. Fall back to task's stored config if state unavailable
3. Only default to 0 if neither source has a config

Author: roomote

src/core/task/Task.ts

@@ -3640,7 +3640,7 @@ export class Task extends EventEmitter<TaskEvents> implements TaskLike {
 		if (Task.lastGlobalApiRequestTime) {
 			const now = performance.now()
 			const timeSinceLastRequest = now - Task.lastGlobalApiRequestTime
-			const rateLimit = apiConfiguration?.rateLimitSeconds || 0
+			const rateLimit = (apiConfiguration ?? this.apiConfiguration)?.rateLimitSeconds || 0
 			rateLimitDelay = Math.ceil(Math.min(rateLimit, Math.max(0, rateLimit * 1000 - timeSinceLastRequest) / 1000))
 		}
 
@@ -3971,7 +3971,7 @@ export class Task extends EventEmitter<TaskEvents> implements TaskLike {
 
 			// Respect provider rate limit window
 			let rateLimitDelay = 0
-			const rateLimit = state?.apiConfiguration?.rateLimitSeconds || 0
+			const rateLimit = (state?.apiConfiguration ?? this.apiConfiguration)?.rateLimitSeconds || 0
 			if (Task.lastGlobalApiRequestTime && rateLimit > 0) {
 				const elapsed = performance.now() - Task.lastGlobalApiRequestTime
 				rateLimitDelay = Math.ceil(Math.min(rateLimit, Math.max(0, rateLimit * 1000 - elapsed) / 1000))