Support shots for expectation value

[sansiro77]

• shots can be set via initialization, measure, and expectation.
• If expectation specifies shots, the estimated value obtained by sampling will be returned.
  For example:

data = torch.randn(4)
cir = dq.QubitCircuit(4, shots=2048)
cir.rylayer(encode=True)
cir.cnot_ring()
cir.rxlayer()
cir.observable(0, 'y')
cir.observable([0,1], 'xy')
cir.observable([0,1,3], 'xyz')
state = cir.amplitude_encoding(torch.randn(2**4))
cir(data=data, state=state)
# cir(data)
print(cir.shots)
print(cir.measure())
print(cir.expectation())
print(cir.expectation(shots=10000))
print(cir.shots)