No timeout on any outbound request

Summary

All three outbound HTTP calls use bare fetch with no AbortSignal. A stalled connection hangs the tool call indefinitely, and a stdio MCP server has no other liveness mechanism, so the client is left waiting with no error and no way to tell the difference between slow and dead.

Locations

• src/auth/meetup-token-provider.ts:66 (token exchange)
• src/meetup/graphql-client.ts:46 (all reads and writes)
• src/meetup/image-service.ts:78 (image upload)

Suggested fix

Add signal: AbortSignal.timeout(30_000) to each call, and map the resulting TimeoutError onto the existing network error category. The write paths already have good "this may have succeeded, inspect Meetup before retrying" wording that should apply to a timeout too.

Start with the three fetch locations: src/auth/meetup-token-provider.ts:66, src/meetup/graphql-client.ts:46, and src/meetup/image-service.ts:78. Read the existing network error handling and the write-path retry wording before updating the calls. Done means stalled requests time out after 30 seconds, timeouts map to the network category, and write operations retain the existing inspection-before-retry guidance.