fix: handle null metadata on multicircuit parent job responses

[splch]

Summary

• The IonQ API returns metadata: null and empty stats: {} on parent multicircuit job responses when the job was submitted without qiskit metadata (e.g. via the REST API directly, or any non-qiskit client). This caused IonQJob.status() to crash with AttributeError: 'NoneType' object has no attribute 'get' when retrieving such jobs.
• Three null-safety fixes in ionq_job.py, plus a fallback that infers qubit count from result data when metadata is entirely absent.
• Verified against the live IonQ API (v0.4 simulator) with both qiskit-submitted and raw-submitted multicircuit jobs.

Root cause

response.get("metadata", {}).get("qiskit_header") does not guard against a present-but-null value: when the key "metadata" exists with value None, the default {} is not used, and None.get(...) raises AttributeError.

Fixes

1. status() line 395: response.get("metadata", {}) -> (response.get("metadata") or {}) so None becomes {}
2. status() line 400: Guard _meas_map_from_header(h or {}, ...) against None header entries
3. _format_result() line 549: Guard (qiskit_header[i] or {}).get("n_qubits", ...) against None header entries
4. _format_result() qubit inference: When clbits is empty and result data exists, infer qubit count from the highest result key (max(int(k)).bit_length()) so _build_counts gets a usable measurement map

Test plan

• New test test_multicircuit_null_metadata_status - verifies status() doesn't crash with null metadata parent
• New test test_multicircuit_null_metadata_result - verifies end-to-end result() with aggregated results keyed by child job ID
• New fixture dummy_multicircuit_parent_response matching the real IonQ API shape
• All 341 existing tests still pass
• Verified against live IonQ API (v0.4 simulator)