Shuttle schedule serialization should serialize on every panic.

Currently we only serialize the schedule on the first panic.

This is wrong for two reasons:
1. We keep scheduling after the first panic. We should append this scheduling information to get the complete schedule executed.
2. If a panic is ever caught then the panic hook will run and we will serialize, but we will keep going. If the test then has an uncaught panic then the schedule will be of little value.

No file or test is named; start by locating the schedule serialization path and panic hook. Verify that scheduling information is appended after every panic, including caught and later uncaught panics, and that the complete executed schedule is covered by tests.

Written by the indexing model from the issue text.