feat(api): add automatic retry with backoff for Figma API rate limiting

[maroffo]

Summary

Add automatic retry with exponential backoff for Figma API rate limiting.

Context: As of November 17, 2025, Figma has updated their API rate limits. See the official announcement and rate limits documentation.

This PR handles:

• HTTP 429 (Too Many Requests) responses with automatic retry using the Retry-After header
• Request timeouts with exponential backoff (1s → 2s → 4s)
• Graceful exit when Retry-After exceeds 5 minutes (avoids indefinite waiting)

Changes

• Add RateLimitError enum for rate limiting error types
• Add RetryConfiguration struct for customizable retry behavior
• Add HTTPURLResponse extension to extract Retry-After header
• Add requestWithRetry() method to Client protocol and BaseClient
• Update FigmaClient to support retry configuration
• Update all loaders to use requestWithRetry()
• Add FigmaAPITests test target with rate limiting tests

Default Configuration

Setting	Value
Max retries	3
Max Retry-After wait	300s (5 min)
Initial timeout backoff	1s
Backoff multiplier	2.0x

Test Plan

• All existing tests pass (67 tests)
• New rate limiting tests pass (12 tests)
• Builds on macOS

[subdan]

It's a bad practice to use Thread.sleep. It blocks the thread. Try to use Timer or DispatchQueue+asyncAfter

[maroffo]

Good catch! Replaced Thread.sleep with a hybrid RunLoop-based implementation that keeps the thread responsive during the wait. When RunLoop has active sources, it processes pending callbacks; otherwise, it falls back to short Thread.sleep(0.1) intervals to ensure the delay actually elapses (pure RunLoop would return immediately on threads without input sources).

Additionally, I've added a configurable requestDelay parameter (under figma: in YAML) to proactively throttle requests between API calls, which helps prevent rate limiting when exporting large numbers of assets.

Example config:

figma:
  lightFileId: "..."
  requestDelay: 0.1  # 100ms between requests

[alexey1312]

+1