WIP: progressive read timeout review

[BGQ99]

Summary

This draft PR prepares the progressive read timeout changes from apache#3409 for review.

Changes

• Rework progressive read timeout tracking around an idle deadline.
• Refresh the deadline after each successful OnReadOnePart() callback.
• Keep timer callbacks independent from the wrapper and user reader lifetime.
• Close the HTTP/1.x socket through Socket::SetFailed() instead of destroying the parsing context from the timer thread.
• Ensure the final reader callback is delivered at most once.
• Preserve errors returned by the user progressive reader.
• Reject progressive read timeout on HTTP/2 with ENOTSUP.
• Restore SocketId plumbing and initialize _socket_id with INVALID_SOCKET_ID.
• Register EPROGREADTIMEOUT.
• Add timeout, slow-callback, reader-error, HTTP/2, and callback-lifecycle tests.
• Fix the http_c++ example buffer handling, reader cleanup, and spelling.

Verification

• Build completed successfully.
• Progressive timeout tests passed repeatedly.
• HTTP test suite: 52/52 passed.
• git diff --check passed.

Known Test Coverage Gap

The active progressive-read test currently verifies successful continuous reading, but does not yet explicitly guarantee that the total transfer time exceeds the configured timeout while each inter-read interval remains below the timeout.

The production deadline-refresh logic is covered by the current behavior tests. The active-read timing test should be strengthened in a follow-up revision.

This is a draft for colleague review. No merge is intended yet.