[ps2]

Stacked on #28 (ignoreNextMessages). Merge that first, or review only the second commit.

POST /command {"command": "failNextConnects", "value": N, "hangUpAfter": 7} makes the next N connects fail the way some pods do in loopandlearn/OmnipodKit#199:

1. The link comes up.
2. Service and characteristic discovery go unanswered.
3. The pod hangs up after hangUpAfter seconds (default 7) with CBErrorDomain code 7, "The specified device has disconnected from us."

Only -transport impossible supports it. For the app to see code 7 rather than a generic ImpossiBLE error, its ImpossiBLE needs mickeyl/ImpossiBLE#9.

Testing:

• The new Go test checks four things:
  • discovery goes unanswered on a failing connect
  • the hang-up carries domain and code
  • the pod is disconnected afterwards
  • the next connect is served normally
• go test ./pkg/bluetooth passes.
• Ran it against Loop next-dev in the iOS Simulator with an Omnipod DASH. Two failing connects reproduced the #199 pattern:
  • Each connect timed out in service discovery after 2 s.
  • The pod hung up at 7.0 s, and Loop logged CBErrorDomain 7.
  • Loop's next connect was set up normally. With an unpatched ImpossiBLE, it isn't (Release a peripheral before telling the delegate it disconnected mickeyl/ImpossiBLE#8).