[CHERIoT] Improve test convergence through custom instruction sequences, which create valid capabilities instead of quickly clearing the tag through randomization

This issue has no description.

The issue names CHERIoT test convergence, custom instruction sequences, valid capabilities, and randomization but provides no files, tests, or entry points. Start by locating the CHERIoT tests and their randomization logic; done means custom sequences improve convergence while producing valid capabilities.

Written by the indexing model from the issue text.