Floss: Refactor flow when controller needs to reset

Currently cases that controller needs to reset are
- HCI command timeout.
- Receives HCI Hardware error event.
- Fails to write to or read from the socket.

The current possible issues are
- On HCI command timeout or Hardware error
  The HciLayer might still send and receive packets. It becomes more
  complicated for stack to shutdown the stack and handle the packets at
  the same time.

- When socket fails to write or read
  Since HCI Layer is unaware of it, the on_hci_timeout could still be
  fired and invokes additional cleanup procedure.

- In either case
  We send SIGTERM to the process. The Floss stack will initiate a
  shutdown procedure which involves a series of HCI commands, but the
  commands might just fail or timeout because the controller is already
  in a bad state.

The CL does following upon any of the above issue happens
1. Block all incoming packet callbacks and skip to send outgoing packets
2. Use SIGUSR1 signal to represent the signal that the controller needs
   to be reset

NO_IFTTT= Floss only change.

Bug: 393662321
Tag: #floss
Test: mmm packages/modules/Bluetooth
Flag: EXEMPT, Floss-only changes

Change-Id: I3ea011f56943f6632da00bf2dabd9be2110578bb