fix(weixin): add retry mechanism for sendMessage ret=-2 errors

[chenhg5]

Summary

• Add retry mechanism (3 attempts) for sendMessage when ilink API returns ret=-2
• Refresh context_token between retries to handle expired tokens
• Add 100ms delay between message chunks to avoid rate limiting
• Add 500ms delay between retries for temporary failures
• Add more detailed logging including content_len for debugging

Root Cause

When ilink API returns ret=-2, it typically indicates:

• context_token expired or invalid
• temporary rate limiting

Previously, cc-connect would immediately fail without retry, causing partial message delivery.

Changes

• weixin.go: Add sendChunkWithRetry with retry logic and chunk delay
• media_outbound.go: Add sendSingleItemWithRetry for media messages
• client.go: Add content_len to error logging for better debugging

Test plan

• go build ./... passes
• go vet ./... passes
• go test ./platform/weixin/... passes
• Manual test: send long messages that get split into chunks
• Manual test: verify retry works when context_token expires

🤖 Generated with Claude Code

Fixes #365

[chenhg5]

✅ QA review: Ready to merge

Well-structured retry mechanism for weixin API failures.

Review summary:

• ✅ Logic correct: 3 retries with 500ms delay, context_token refresh
• ✅ CI passes (all 5 checks green)
• ✅ Proper context cancellation handling
• ✅ Chunk delay (100ms) to avoid rate limiting
• ✅ Clear logging with attempt numbers
• ✅ Fixes [Bug] 个人微信（weixin）场景下回复已生成但 sendMessage 被拒，导致微信侧回复不完整/后半段丢失 #365

Minor observation: Retry logic duplicates between sendChunkWithRetry and sendSingleItemWithRetry - could consider extracting to a helper in a follow-up PR, but not a blocker.

Good fix for a real production issue.

[chenhg5]

LGTM! Solid retry mechanism for weixin sendMessage failures.

✅ Implementation:

• sendChunkWithRetry and sendSingleItemWithRetry handle ret=-2 errors with 3 retries
• Context token refresh between retries
• 100ms chunk delay and 500ms retry delay for rate limiting

✅ Root cause handling: ret=-2 typically indicates expired context_token or rate limiting - retry with fresh token is the correct approach.

✅ Logging: Added content_len for debugging which is helpful.

✅ CI: All checks pass.

Ready for merge.