Bluetooth: hci_uart broken on master

[carlescufi]

Today we bisected an issue with HCI events coming from the hci_uart controller being corrupted (i.e. spurious data) to c66bb21

it happens on first connection, very easy to reproduce. Just flash hci_uart onto a Nordic board, use QEMU/native posix as Host and then start advertising on the shell, connect from any peer central to it.
It typically happens always after the first connection is established.

This is the sort of thing that you see on btmon:

> HCI Event: LE Meta Event (0x3e) plen 31                                                    #3025 [hci0] 3141.057741
      LE Enhanced Connection Complete (0x0a)
        Status: Success (0x00)
        Handle: 0
        Role: Slave (0x01)
        Peer address type: Public (0x00)
        Peer address: 00:28:F8:B3:4D:08 (Intel Corporate)
        Local resolvable private address: 7E:6C:28:71:4C:C3 (Resolvable)
        Peer resolvable private address: 00:00:00:00:00:00 (Non-Resolvable)
        Connection interval: 56.25 msec (0x002d)
        Connection latency: 0 (0x0000)
        Supervision timeout: 9600 msec (0x03c0)
        Master clock accuracy: 0x01
< HCI Command: LE Read Remote Used Features (0x08|0x0016) plen 2                             #3026 [hci0] 3141.065027
        Handle: 0
> HCI Event: Connect Request (0x04) plen 15                                                  #3027 [hci0] 3141.139494
        invalid packet size
        04 00 01 16 20 04 04 3e 0c 04 00 00 00 ff 00     .... ..>.......