Rework flint startup

Closes #2582 (closed)

This MR rework the way Flint starting in order to be a bit stronger versus stucked state.

• Re-implement the low API which create/attach/close Flint
• Better handling of timeout. We don't want 30s timeout on user interaction
• A new argument to get_flint(): restart_if_stuck, False by default
• A new property to SCAN_DISPLAY: restart_flint_if_stuck, False by default (used when auto=True)

This also changes the ontology of the Flint proxy inside BLISS.

Previously there was a new proxy for each Flint, which was dangerous, cause it was not possible to hold a Flint proxy without having trouble when Flint was recreated. Now it is handled as a singleton. The Flint proxy is unique, and can be connected to different Flint. This could be used to provide BLISS side setup when a new Flint is created/connected.

This feature is used on test fixtures, to track Flint PID and clean up the env at the end of each tests.