[rom,rom_ext] Test sideload-clear timing for keymgr.

### Description

See https://github.com/lowRISC/opentitan/pull/20011#discussion_r1357813813

Currently we spin for 100 microseconds before stopping keymgr clearing; we should thoroughly test to make sure this is enough, and possibly shorten it to save time if it's more than enough. An alternative would be to leave the clearing on until the next time we need to sideload something, but that could consume a lot of entropy unnecessarily.

Start with the linked pull request discussion and trace the keymgr sideload-clear behavior around the current 100-microsecond wait. Determine how to test whether clearing completes sooner without weakening the intended protection, then verify that any timing change preserves correct sideload behavior and avoids unnecessary entropy consumption.

Written by the indexing model from the issue text.