docs(workflow-executor): fix README inaccuracies

[Scra3]

What

Fixes three inaccuracies in packages/workflow-executor/README.md, found while writing the developer-guide setup page:

1. POLLING_INTERVAL_MS default — documented as 5000, actually 30000 (defaults.ts: DEFAULT_POLLING_INTERVAL_MS = 30_000). Fixed in both the env table and the log example.
2. Startup log example — was missing the Agent probe passed line (logged in runner.start()) and malformed=0 on the Poll cycle completed line (third context key in runner.ts). The poll timestamp was also wrong: the first poll fires after the interval (setTimeout(..., pollingIntervalMs)), not ~5s in.
3. Graceful shutdown / exit codes — the README claimed steps are "force-killed and the process exits with code 1" after STOP_TIMEOUT_MS. In reality (runner.ts Promise.race + build-workflow-executor.ts), a drain timeout logs a Drain timeout error, abandons the in-flight steps, and still exits 0. Exit 1 only happens on startup error or a resource-cleanup failure during shutdown.

Why

The developer-guide page initially inherited these errors from the README. Fixing the source so the next person doesn't recopy them.

🤖 Generated with Claude Code

Note

Fix inaccuracies in workflow-executor README

Updates README.md to correct several inaccuracies introduced by prior changes.

• Changes documented POLLING_INTERVAL_MS default from 5000 to 30000 and updates example timestamps accordingly
• Adds an "Agent probe passed" log line to the startup example and a "malformed" counter to poll cycle log examples
• Revises the graceful shutdown description: polling stops first, steps still running after STOP_TIMEOUT_MS are not awaited or force-killed, a "Drain timeout" is logged, and interrupted steps are safe to replay via a write-ahead idempotency log
• Corrects exit codes: code 0 now covers graceful shutdown even when drain timed out; code 1 now covers startup errors or resource cleanup failures

^{Macroscope summarized 25d3dc5.}