Proper handling of serialized numpy metadata

[coruscating]

FineAmplitude fails when a numpy array is passed in as an experiment option:

qc = QuantumCircuit(2)
qc.rzx(np.pi/2,0,1)

with pulse.build(backend=backend, name="szx") as rzx:
    pulse.call(qc)

inst_map = InstructionScheduleMap()
inst_map.add("szx", (0,1), rzx)

fine_amp = FineZXAmplitude((0,1), backend)
fine_amp.set_transpile_options(inst_map=inst_map)
fine_amp.set_experiment_options(repetitions=np.arange(0,10,1))

expdata = fine_amp.run(backend).block_for_results()

This code doesn't run at all with qiskit-ibm-provider < 0.6.1, which is the version that fixes numpy serialization (we should probably add this as a requirement). With 0.6.1, the error occurs in the analysis:

Traceback (most recent call last):
  File "qiskit_experiments/framework/experiment_data.py", line 897, in _run_analysis_callback
    callback(self, **kwargs)
  File "qiskit_experiments/framework/base_analysis.py", line 169, in run_analysis
    results, figures = analysis._run_analysis(expdata)
  File "qiskit_experiments/curve_analysis/curve_analysis.py", line 384, in _run_analysis
    processed_data = self._run_data_processing(
  File "qiskit_experiments/curve_analysis/curve_analysis.py", line 172, in _run_data_processing
    xdata = np.asarray([datum["metadata"][x_key] for datum in analyzed_data], dtype=float)
TypeError: float() argument must be a string or a number, not 'dict'

This happens because these numpy ints are serialized to {"__type__": "int", "__value__": int(obj)} dicts in the provider. @nkanazawa1989 Perhaps just adding handling for these types in CurveAnalysis will fix this?